ICL RM OM Clit reLU by Tia Applications to Robotics JOSEPH DUFFYRobot manipulators are becoming increasingly important in research and in- dustry, and an understanding of statics and kinematics is essential to solving problems of robotics. This book provides a thorough introduction to statics and first-order in- stantaneous kinematics with applications to robotics. The emphasis is on se- rial and parallel planar manipulators and mechanisms. The text differs from others in that it is based solely upon the concepts of classical geometry. It is the first to describe how to introduce linear springs into the connectors of parallel manipulators and to provide a proper geometric method for controlling the force and motion of the hands (end effectors) or tools of serial robot manipulators performing constrained motion tasks. Both students and practicing engineers will find this book easy to follow, with a clearly written text and abundant illustrations, as well as exercises and real-world projects to work on. Professor Joseph Duffy is Director of the Center for Intelligent Machines and Robotics at the University of Florida.Statics and kinematics with applications to roboticsStatics and kinematics with applications to robotics JOSEPH DUFFY University of Florida CAMBRIDGE UNIVERSITY PRESSCAMBRIDGE UNIVERSITY PRESS Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo Cambridge University Press The Edinburgh Building, Cambridge CB2 2RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521482134 © Cambridge University Press 1996 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 1996 This digitally printed first paperback version 2006 A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Duffy, Joseph, 1937— Statics and kinematics with applications to robotics / Joseph Duffy. p. cm. Includes index. ISBN 0-521-48213-5 (hardcover) 1, Manipulators (Mechanism) 2. Statics. 3. Machinery, Kinematics of. 4, Robots - Motion. I. Title. TI211.D84 1996 95-4913 670.42'72 ~ de20 cp ISBN-13 978-0-521-48213-4 hardback ISBN-10 0-521-48213-5 hardback ISBN-13 978-0-521-03398-5 paperback ISBN-10 0-521-03398-5 paperbackTo my wife, AnneContents 11 1.2 13 16 Preface Introduction A displacement analysis of planar robot manipulators Mechanisms and manipulators The mobility of planar mechanisms and manipulators Displacement analysis of single degree of freedom (d.o-f.) planar manipulators Displacement analysis of two-link serial manipulators 1.4.1 Analysis of the RR manipulator 1.4.2 Analysis of the RP manipulator 1.4.3 Analysis of the PR manipulator 1.4.4 Analysis of the PP manipulator Exercise 1.1 Analysis of 3R and 2R-P three-link serial manipulators 1.5.1 Analysis of the RRR manipulator 1.5.2 Analysis of the RRP manipulator 1.5.3 Analysis of the RPR manipulator 1.5.4 Analysis of the PRR manipulator Exercise 1.2 Analysis of 2P-R three-link serial manipulators vii page xi 13 14 19 20 21 21 23 23 24 27 28 30 31 32Contents viii 1.7. Analysis of redundant-serial manipulators with four or more kinematic pairs 33 1.8 Displacement analysis of in-parallel manipulators 34 1.8.1 Description of in-parallel manipulators 34 1.8.2 A reverse analysis for the general in-parallel manipulator 36 1.8.3. A forward analysis for an in-parallel manipulator with the simplest geometry 38 Exercise 1.3 39 2 Planar statics 40 2.1 The coordinates of a line in the XY plane 40 2.2 The coordinates for the point of intersection of a pair of lines 48 Exercise 2.1 49 2.3 The statics of plane rigid systems 49 2.4 Parallel translation and rotation of rectangular coordinate systems 56 2.4.1 Parallel translation of a rectangular coordinate system 56 2.4.2 Rotation of a rectangular coordinate system 60 2.4.3 The Euclidean group of motions 63 2.5 Induced force/line transformation under the Euclidean group 64 Exercise 2.2 65 2.6 A useful vector representation of a force 66 2.7 The statics of a parallel manipulator 67 Exercise 2.3 7 2.8 The geometrical meaning of j~! 72 2.9 Singularity configurations of a parallel manipulator 15 Exercise 2.4 719 2.10 Statically redundant parallel manipulators 80 3 First-order instantaneous planar kinematics 82 3.1 The definition of a rotor 82 3.2 The coordinates of a line parallel to the Z axis 83 3.3. A useful vector representation of a rotor 85Contents ix 3.4 Infinitesimal displacements of a rigid lamina 87 3.5 A representation of pure translation 88 3.6 Ray and axis coordinates of a line and a rotor 89 3.7 Translation and rotation of coordinate systems 91 3.8 Induced line (rotor) transformation of the Euclidean group 93 3.9 Relationship between [e] and [E] 93 Exercise 3.1 94 3.10 The first-order instantaneous kinematics of a 3R serial manipulator 94 3.11 The geometrical meaning of J~! 97 Exercise 3.2 99 3.12 Singularity configurations of a serial manipulator 102 Exercise 3.3 103 3.13 Kinematics of redundant serial manipulators 106 3.14 First-order instantaneous kinematics of a closed-loop 4R mechanism 107 Exercise 3.4 110 3.15 Instantaneous kinematics of a 3R serial manipulator using displacement equations 112 4 Dualities of parallel and serial devices 117 4.1 Introduction 117 4.2 Mutual moment, instantaneous power, and instantaneous work 118 Exercise 4.1 127 4.3. A static analysis of a planar serial manipulator 129 Exercise 4.2 134 4.4 A static analysis of a parallel manipulator with RPR and 3R kinematic chains 136 Exercise 4.3 139 4.5 A kinematic analysis of a parallel manipulator 140 4.6 Instantaneous kinematics of parallel manipulators in singularity configurations 146 4.7 An infinitesimal displacement analysis for a parallel manipulator 148 4.8 The differential of a line 151Contents x 5 The stiffness mapping for a parallel manipulator 153 5.1 A derivation of the stiffness mapping 153 5.2. The dimensions of the elements of the stiffness matrix 157 5.3 The stiffness mapping of a planar two-spring system 159 5.4 Force and motion control using a serial manipulator with a compliant wrist 164 5.5 A note on the stability of spring systems 168 Exercise 5.1 168 References 171 Index 173Preface This text is devoted to the statics of rigid laminas on a plane and to the first-order instantaneous kinematics (velocities) of rigid laminas moving over a plane. Higher-order instantaneous kinematic problems, which involve the study of accelerations (second-order properties) and jerk (third-order prop- erties) are not considered. This text is influenced by the book Elementary Mathematics from an Ad- vanced Standpoint: Geometry, written by the famous German geometer Fe- lix Klein. It was published in German in 1908 and the third edition was trans- lated into English and published in New York by the Macmillan Company in 1939. The book was part of a course of lectures given to German High School Teachers at Géttingen in 1908. Klein was admonishing the teachers for not teaching geometry correctly, and he was essentially providing a proper foundation for its instruction. The present text stems from the undergraduate course “The Geometry of Robot Manipulators,” taught in the Mechanical Engineering Department at the University of Florida. This course is based on Klein’s development of the geometry of points and lines in the plane and upon his elegant development of mechanics: “A directed line-segment represents a force applied to a rigid body. A free plane-segment, represented by a parallelogram of definite con- tour sense, and the couple given by two opposite sides of the parallelogram, with arrows directed opposite to that sense, are geometrically equivalent con- figurations, i.e., they have equal components with reference to every coordi- nate system.” xiPreface xii It is the author’s opinion that a student’s understanding of statics and in- stantaneous kinematics is enhanced immeasurably by learning that quantities such as force and turning moment have a geometrical meaning, and that they are equivalent to a directed line segment and a directed area, respectively. This facilitates understanding the nature of invariance and explains how the behavior of such quantities is influenced by the group of Euclidean trans- formations. It is sad that the great developments in geometry of the last century and its application in mechanics have, for the most part, been forgotten or ig- nored by many researchers in the field of robotics. Perhaps the most blatant exclusion is that no reference is made by modern researchers in robotics to the theory of constrained motion, which was completed by Sir Robert Stawell Ball in his monumental Treatise on the Theory of Screws (1900, Cambridge University Press). Because of this omission, a large number of erroneous ar- ticles have been presented at national and international robotics conferences and in the technical journals of learned societies. These articles aim to model constrained-motion tasks, grasping operations, and the calibration of robots. Ball correctly modeled the constrained motion of a rigid body by permissi- ble, instantaneous rigid-body motions together with constraint forces/couples in combinations, which he defined well over a century ago as, respectively, twists of freedom and wrenches of constraint. Many researchers either are unaware of Ball’s work and/or choose to ig- nore it. They introduce fallacious concepts, such as “the motions a constrained body cannot have,” which they bravely define as “natural constraints.” It con- cerns me to see these errors enshrined in a growing number of textbooks on robotics for engineering students. Such texts subject students to meaningless expressions in which dimensions and units are mixed, for example, (Linear Velocity)? + (Angular Velocity), and present answers to problems in which the values of physical quantities change with the choice of reference point (i.e., the answers are not invariant with a translation of origin). Students are also subjected to least-squares minimization problems with objective func- tions such as (a? + 6), where d and 6, respectively, have units of length and radians. These incorrect concepts are devoid of geometric meaning, and they can have no physical meaning either. As we advance toward the 21st cen- tury, the rigorous study of statics and instantaneous kinematics applied to thePreface xiii field of robotics is apparently receding and being regarded as archaic. The author hopes that this short text will help to reverse this trend. The author wishes to thank Dr. E. J. F. Primrose (formerly Mathematics De- partment, University of Leicester), Dr. J. M. Rico-Martinez (Instituto Tec- nolégico de Celaya), and Dr. M. G. Mohamed (El-Mina University) for their valuable comments and suggestions on the text. The author also wishes to thank Dr. M. Griffis, Mr. I. Baiges, Mr. R. Hines, and Mr. S. A. Kelkar (Uni- versity of Florida) for their contributions to the text. The author is indebted to Ms. Cindy Townsend for typing the manuscript, and to Ms. Florence Pad- gett, Editor, Physical Sciences, Cambridge University Press, for her advice and guidance in the preparation of the text. J. Duffy Gainesville, Florida October 1995Introduction Statics and first-order instantaneous kinematics are intimately related. They are completely analogous or, more specifically, they are dual concepts as, for example, a line and a point are dual in the projective plane. The meet of two lines is always a point in the projective plane, on which parallel lines are said to meet at a point at infinity. This proposition can be re-stated for two points by making the appropriate grammatical changes in order for it to make sense. The dual statement is simply that the join of two points is a line. It is always possible to formulate (prove) a proposition (theorem) for one dual el- ement and to simply state a corresponding proposition (theorem) for the cor- responding dual element. Another example for lines and points is: Two distinct lines which have Any two distinct points acommon point determine determine a range of points on a planar pencil of lines through the line which joins the that point. two points. In statics a line segment represents the “linear” concepts of force, whereas in instantaneous kinematics a line segment represents the “circular” concept of instantaneous rotation. This book evolved from lecture notes for a first course in robot manipu- Jators, an undergraduate course in mechanical engineering at the University of Florida, which was first offered in the fall of 1989. The major objective of the course is to provide students with a proper understanding of statics and first-order instantaneous kinematics. The prerequisites are that students 1Introduction 2 have passed entry level courses in statics and dynamics and that they have a working knowledge of a computer language. This text is primarily (Chapters 2-5) devoted to a study of planar statics and first-order instantaneous kinematics with applications to manipulators. However, to perform a static and instantaneous kinematic analysis of a ma- nipulator it is necessary to know the geometry of the manipulators. That is, we need to know the relative angles between consecutive pairs of links con- nected by turning joints and the relative linear displacements between pairs of consecutive links connected by sliding joints. An analysis of the geome- try of planar manipulators is given in Chapter 1. The elements of mechanisms and manipulators are also introduced in Chapter 1. This is followed by determining the so-called forward and reverse position analysis of planar, serial two- and three-jointed manipulators. At the outset, the student is required to draw a planar serial manipulator, and to iden- tify and label joint variables with manipulator dimensions. Equations are then obtained by parallel projection. A forward analysis is relatively straightfor- ward. A unique position for a point in the end effector together with the end- effector orientation can be computed for a specified set of joint variables. The same equations are used to perform a reverse analysis; the coordi- nates of a point in the end effector are specified and usually two configura- tions (elbow up and elbow down) are computed for a two-jointed serial ma- nipulator. Analogously, the coordinates of a point in the end effector together with the end-effector orientation for a serial three-jointed manipulator are spec- ified, and a pair of elbow up and elbow down configurations are computed. Chapter 1 concludes with a reverse analysis of a parallel manipulator with the general geometry for which a rigid lamina (called a moving platform) is connected to the ground via three revolute-slider-revolute kinematic chains, typically called connectors. Here the reverse solution is relatively straight- forward. The specification of the location (position and orientation) of the moving platform yields a unique solution for the three connector lengths. However, specifying the three connector lengths leads to multiple forward solutions for the location of the moving platform. For the most general case, there may be as many as six locations, which is complicated. Finally, an ex- ample of a parallel manipulator with a special geometry that yields four lo- cations of the moving platform is given. Chapter 2 develops the planar statics of rigid laminas based upon the firmIntroduction 3 foundation of geometry. It follows closely the development of F. Klein, and forces and moments are shown to be geometrically equivalent to directed line segments and areas, respectively. A force can be considered a scalar multi- ple of a line, and homogeneous line coordinates are synonymous with force coordinates. A study of a system of forces is essentially a study of a system of lines. The resultant force that acts upon the platform of a parallel device is a lin- ear combination of the system of forces generated in the connectors of the platform (forward static analysis). Conversely, an external force that acts upon the platform will produce a component resultant in each platform connector (reverse static analysis). Chapter 3 develops the planar, first-order, instantaneous kinematics of lam- inas. As in Chapter 2, this chapter follows closely the development of F. Klein. An instantaneous rotation of a rigid lamina is essentially a scalar mul- tiple of a line, which is the rotation axis. A study of the relative instantaneous motion of a number of serially con- nected laminas is really a study of the lines which define the various axes of rotation. The instantaneous motion of the end effector of a serial planar ma- nipulator is a linear combination of the rotation speeds of revolute joints or the linear displacement speeds of sliding joints in the chain (forward kinematic analysis). Conversely, when the instantaneous motion of the end effector of a serial manipulator is specified, the joint motions required to produce the re- quired end-effector motion can be calculated (reverse kinematic analysis). It should become clear to the student that not only are statics and kinemat- ics analogous, or dual, but the statics of a parallel manipulator is completely analogous, or dual, with the instantaneous kinematics of a serial manipulator. That is, the resultant force that acts upon the end effector (the resultant instant motion of the end effector) of a parallel manipulator (of a serial manipulator) is a linear combination of the connector forces (instant joint motions). Fur- thermore, the forward and reverse analyses of the statics and instantaneous kinematics of parallel and serial manipulators are shown to be dual. Chapter 4 establishes the complete duality of statics and kinematics, and shows that the forward and reverse statics analyses of a serial device and the forward and reverse instantaneous kinematics of a parallel device are also dual. Again, as in the previous chapters, this chapter is based on a firm foundation of geometry by introducing such invariant quantities as instantaneous powerIntroduction 4 and instant work, which, for the instantaneous rotation of a lamina produced by a force in the plane of motion, are scalar multiples of the mutual moment of a pair of lines, the rotor axis, and the line of action of the force. The concepts of the mutual moment of a pair of lines, instant power, and instant work date back well over a century, as do the conditions for the van- ishing of instant power (or work) discovered by Sir Robert Stawell Ball. The invariant condition for the vanishing of instantaneous power was defined by Ball as reciprocity, and he based his theory of constraints and freedoms of partially constrained rigid bodies on this definition. These are the same con- cepts that have largely been ignored by many modern researchers in robot- ics. A new theory, called the modern hybrid control theory, for the control of force and motion has surfaced. This theory is based on a condition that is not invariant with a choice of reference point. It yields equations with mixed units, has no geometrical meaning whatsoever, and is therefore nonsense. Chapter 4 concludes with an infinitesimal displacement analysis of a par- allel manipulator together with the derivative of a line rotating in a plane about an axis normal to the plane of rotation. These expressions are used in Chapter 5 to compute a relatively new stiffness mapping, which relates a small change of force acting upon a platform of a parallel manipulator, ini- tially in static equilibrium, to the instantaneous rotation of the platform pro- duced by the change of force. The spring matrix is a function of the spring constants of linear springs introduced into the legs together with the preload and the geometry. In 1965 F. M. Dimentberg derived a spring matrix for a parallel manipulator which remains close to its unloaded configuration. It is well known that a small change in force 5fcan be produced by a small displacement dx of a linear spring, and 6f = kx, where k is the spring con- stant. This means that a small change in force could be controlled by the small displacement of a spring. Chapter 5 extends this concept to simultane- ously control the force and the displacement of a point connected to the ground via a pair of linear springs. Following this, it is extended to simulta- neously control the force and motion of a rigid lamina, the platform of a par- allel manipulator connected to the ground via three compliant revolute-slider- revolute connectors. A linear spring is inserted in each slider joint. A variety of numerical examples are given throughout the text. It is ad- vantageous for students to have access to computer graphic facilities so they can animate manipulators and display their results.1 A displacement analysis of planar robot manipulators 1.1. Mechanisms and manipulators A rigid lamina is free to move on a fixed reference plane, the page. The lamina is unbounded and covers the entire reference plane. However, for ease of visualization an arbitrary contour is drawn on the moving lamina, and the moving lamina is usually identified with the area enclosed by the con- tour. The motion of the lamina measured relative to the fixed plane can be de- termined by first specifying some reference point O in the fixed plane. In Figure 1.1, O is chosen outside the contour (it could equally be chosen on or inside the contour boundary). Next, some reference point Q on the moving lamina is chosen; its position relative to O can be measured, for instance, by the coordinates (xg, yg). Finally, a reference line is drawn on the moving lam- ina, and its orientation y is measured relative to some reference line in the fixed plane, for example, the X axis. Changes {Axg, Ayg; Ay} in the three parameters {xg, yg; y}* can be used to quantify the motion of the moving lamina. Since each parameter can be varied independently, the moving plane is said to possess three freedoms or three degrees of freedom (3 d.o.f.) with respect to the fixed plane. It is important to recognize that there are many representations of the lo- cation and, hence, the motion of a moving lamina. One is free to select the * The units for y and (xg, yg) are, respectively, radians or degrees and length (inches, feet, or meters, etc.). The semicolon is introduced to indicate this difference in units. 5Displacement analysis of planar robot manipulators 6 Figure 1.1 The location of a lamina in a fixed reference plane. two reference points O and Q together with the orientation of the XY coor- dinate system. The moving lamina may not possess three d.o.f. It may be constrained, for example, to rotate about an axis Z, normal to the reference plane, by a rev- olute R kinematic pair, hinge, or turning joint (see Fig. 1.2). Clearly, all points on the moving lamina move on concentric circles centered on the axis of ro- tation. Here, it is convenient to draw reference lines through point O on the axis of rotation and to measure the motion by changes in the single parame- ter 6. A revolute pair permits one freedom (f = 1) of the moving lamina rel- ative to the fixed plane. Mechanisms and robot manipulators are illustrated by closed and open polygons that use such reference lines. These skeletal forms are essentially geometrical models which can be labeled conveniently with the joint vari- ables and the link lengths. The moving lamina may be constrained to pure translational motion by a prismatic P kinematic pair (sliding joint). A prismatic pair has no special line because all points on the moving lamina translate in a specific direction. It is, however, common practice to designate the center line of the joint as a refer- ence line and to measure the displacement of some point Q on the moving lam- ina relative to some point O in the fixed plane which lies on this reference line. An assemblage of links and joints such as those illustrated in Figure 1.2 was defined by Reuleaux (1876) as a kinematic chain that may be open or closed.Mechanisms and manipulators a, Revolute(R) Representation Figure 1.2 Revolute and prismatic kinematic pairs. It can be a single open or closed loop, or it can be a combination of open and closed loops. Reuleaux stated: “In itself a kinematic chain does not pos- tulate any definite absolute displacement.* One must hold fast or fix in po- sition one link of the chain relatively to the portion of surrounding space as- sumed to be stationary. The relative displacements of links then become absolute. A closed kinematic chain of which one link is made stationary is called a mechanism.” The link that is held fixed is called the frame or the frame of reference. A change in the selection of a reference frame is known as kinematic inversion. *The term “displacement” has been substituted for the term “motion” used in Reuleaux’s original text. The motion of a rigid lamina relative to a reference frame im- plies not only displacement but velocity, acceleration, and so on (see Hunt 1978).Displacement analysis of planar robot manipulators 8 Figure 1.3 Finite rotation pole. The relative displacement between any pair of links is independent of the choice of reference frame. On the contrary, the absolute displacement of links is dependent on the choice of reference frame. A link 23 (see Fig. 1.3) can be displaced to a second position 2'3’ by ro- tating it about a point P, commonly called the pole, fixed in the reference frame, the page (the ground), which is the point of intersection of the per- pendicular bisectors of 22' and 33’. Point P can be modeled by a revolute pair that connects to the ground the moving lamina which contains link 23. This same displacement can be achieved by connecting points 2 and 3, re- spectively, to any pair of points 1 and 4 which lie on the perpendicular bi- sectors S and T of 22’ and 33’ (see Fig. 1.4). Revolute joints are now located at the four vertices of the quadrilateral and, in this way, a 4R mechanism is formed. Link 41 is stationary and link 23 (the coupler) undergoes an anti- clockwise rotation Ad about P when an actuator drives either link 12 or 43 in an anticlockwise direction. Assume that link 23 is held fixed, and it is thus considered the frame of reference. The same relative displacement of the four links of the mechanism can be obtained by rotating link 41 clockwise about P, as shown in Figure 1.5. The quadrilaterals 4123’ and 4'1’23 are congruent. However, the ab- solute displacements of the links are different. Assume that the links of the 3R open chain 1234 are given successive an- ticlockwise angular displacements of 30, 40, and 90 degrees beginning with link 12 connected to the ground via the revolute pair (see Fig. 1.6). Then sup- pose that link 34 is connected to the ground via a revolute joint, and the links are given successive clockwise displacements of 90, 40, and 30 degrees, be-Mechanisms and manipulators 9 Figure 1.4 A 4R planar mechanism. Figure 1.5 A kinematic inversion of the 4R planar mechanism. ginning with link 34. Itis clear from the figure that the relative displacements of the links for the two cases are the same, and the chains 12’3'4’ and 1"2"3"4 are congruent. However, the absolute displacements of the links are different. The closed-loop kinematic chain 4123 with link 41 as the frame of reference is usually called a mechanism. There is one driving link 12 (the crank), and the movement of the mechanism is repeated through each revolution of the crank. The open kinematic chain 1234 with link 12 connected to the ground via a revolute is an example of a serial robot manipulator where an end effector or some form of gripping device is attached at the free end. Each revolute joint is actuated so that the relative displacements of link 12 to the ground,Displacement analysis of planar robot manipulators 10 I 2 is Figure 1.6 Kinematic inversions of a planar 3R open chain. 23 to 12, and 34 to 23 are controlled independently. This device is capable of performing a multitude of tasks in contrast to the previous mechanism, which is capable of only cyclic movement. Robot manipulators are not necessarily serial. Figure 1.7 illustrates two versions of a planar manipulator with a movable triangle 123 connected to the ground (a fixed triangle 1’, 2’, 3’) by three RPR chains in parallel. This device is called an in-parallel robot manipulator, and it is capable of per- forming a multitude of tasks by actuating each of the three prismatic pairs or, alternatively, each of the three grounded revolute pairs.Mobility of planar mechanisms and manipulators 11 Figure 1.7 Examples of in-parallel manipulators. 1.2 The mobility of planar mechanisms and manipulators The closed planar quadrilateral (see Fig. 1.4), the open serial chain (see Fig. 1.6), and the parallel assemblages (see Fig. 1.7) have been given mobility by introducing revolute and prismatic pairs between the rigid bod- ies that form the assemblages. We may now derive the mobility criterion for an assemblage of rigid links and kinematic pairs. We have established that the rigid lamina (or link), which is free to move on the XY plane (see Fig. 1.1), possesses three degrees of freedom. Clearly, if there were n such unconnected links lying on the XY plane the assemblageDisplacement analysis of planar robot manipulators 12 would possess 3n independent degrees of freedom. Also, if one link were fixed in the XY plane, the assemblage would possess 3(n — 1) freedoms. As- sume that any pair of links is connected by at most one kinematic pair with freedom f. Then the relative freedoms between the pair of links is reduced by (3 — f). Therefore, the degrees of freedom of the n links connected by j kinematic pairs are reduced to Se (3 — fj), where i denotes the ith kine- matic pair. The overall number of freedoms of the assemblage are commonly defined as its mobility M, and M=30-)-S 6-f. ay Equation 1.1 is derived without any consideration of the special geome- try of the assemblage; it cannot be applied without modification to assem- blages which, for example, contain two or more prismatic joints that line up, because this duplicates the sliding motion of a single prismatic joint. When the arrangement of pairs is general these results are obtained: (i) For a closed single-loop assemblage the number of pairs equals the num- ber of links, j = n, and (1.1) reduces to M= & fi-3. (1.2) For all f; = 1 and for n = 3, 4,5...,M=0,1,2..., ie. a triangle, a quadrilateral, and a pentagon . . . possess, respectively, zero (a structure), one, and two freedoms. When the quadrilateral has one link fixed to the ground (see Fig. 1.4) it requires a single input drive to control the motion, which is Tepetitive and cyclic. This assemblage is usually called a four-bar mecha- nism. If one link of a hinged pentagon were fixed to the ground it would re- quire two input drives to control the motion. (ii) For an open serial chain the first pair is connected to ground while the end link is free to move in the XY plane. For this case, n = 1 (ground) + n, (the number of moving links in the chain). Here n, = j and thus n = 1 + j. Equation 1.1 reduces to M=>h (13)Single degree of freedom planar manipulators 13 Therefore, serial chains with one end connected to the ground, with all f; = 1 and j(= n,) = 1, 2,3,4..., possess mobility one, two, three, four. ... Such assemblages are employed as manipulators. An end effector is attached at the free end. An end effector is capable of a variety of motions and of perform- ing a multitude of tasks as opposed to the closed-loop, four-bar mechanism, which is only capable of a single cyclic motion. However, note that a single input drive is required to actuate a four-bar mechanism, whereas a drive is required to actuate each joint in a serial chain. Finally, it is interesting to determine the mobility M of the parallel as- semblage illustrated by Figure 1.7. Clearly, there are three RPR chains that connect the base to a movable lamina and therefore j = 9, Furthermore, there are two links in each of the three chains that connect the base to the mov- able lamina and hence n = 8. The substitution of these values in (1.1) yields M = 3(8 — 1)—9 X 2= 21 — 18 = 3. It follows that three input drives must be employed to control the motion of the movable lamina. Either elec- tric motors can be used to actuate the three base revolute pairs or pneumatic (hydraulic) actuators installed to actuate the three sliding joints in each RPR chain. 1.3 Displacement analysis of single degree of freedom (d.o.f.) planar manipulators The assemblages illustrated in Figure 1.8 can be considered one d.o.f. planar manipulators, where the moving lamina is a link connected to the ground by a revolute pair actuated by a motor or a prismatic pair actuated by a hydraulic or pneumatic cylinder. At the free end, there is a gripper or hand on which a reference point Q is selected (Q could also designate a point on a tool held by the gripper). The link length is known, and the single variables @ and S are, respectively, controlled and the corresponding gripper position is Q (a cos 6, a sin 6) and Q(S, a), where a is the distance of Q from the cen- ter line of the joint. The gripper itself may be actuated, but such motions are independent of the gross motion of the robot manipulators and are of no con- cern here. Note that the skeletal form of link a, which connects the tool point Q to the prismatic joint, is drawn perpendicular to the slider displacement S, and its length is measured from the center line of the joint, Q. It could be drawn at any angle a. Clearly, this cannot affect the relative sliding motion, and theDisplacement analysis of planar robot manipulators 14 a’ sina. @' cos. 3 a Figure 1.8 Revolute and prismatic manipulators. coordinates for the new tool point Q' are simply (S + a’ cos a, a’ sin a + b). From here on, links that connect prismatic joints will be drawn perpen- dicular to the direction of sliding. 1.4 Displacement analysis of two-link serial manipulators It is convenient to describe serial manipulators by their sequence of kinematic pairs, beginning with the first or grounded kinematic pair. Figure 1.9 illustrates the four, two-link, two d.o.f. RR, RP, PR, and PP manipula- tors. Here, both kinematic pairs in each serial chain are actuated. The rela- tive angular displacement between a pair of links aj; and ay (ij = 01 (ground), 12; jk = 12, 23 ...) is denoted by 6; and variable slider displacements are labeled S;, j = 1, 2.... The reader will recall (see Fig. 1.8) that the center line of each prismatic pair is designated as a reference line. All link lengths are measured from these center lines. For example, the link length a3 of theDisplacement analysis of two-link serial manipulators — 15 “%y a < IP ay oe» Figure 1.9 Planar two-link manipulators. RP and PR manipulators is measured as the perpendicular distance from the center lines labeled 5. The analysis of two- and three-link manipulators involves the solution of two trigonometrical equations. The first equation is dcos 0=f, (1.4) where the coefficients d and f are known quantities. Provided that |//d| < 1, when f/d > 0, a calculator will display a value &? in the first quadrant (77/2 > 6 > 0) of the unit circle (see Fig. 1.10). SinceDisplacement analysis of planar robot manipulators 16 eo \y -9@ (|\ 7 fld>0 eo x fld<0 T Figure 1.10 Solutions of cos 6 = f/d. the cosine is an even function, a second solution is given by — & or 0 = (2a — &), which is in the fourth quadrant. When f/d < 0, a calculator will display a value @ in the second quadrant (7 > 6 > 7/2) of the unit cir- cle, and a second solution is — &@ or @ = (2% — &), which is in the third quadrant. The pair of values #*) denotes two distinct manipulator configu- rations for a specified end-effector location. It is important to identify and to keep track of these configurations in path planning the end effector. The second equation is esin 0=g, (15) where the coefficients g and e are known quantities. Provided that |g/e| = 1, when g/e > 0 a calculator will display a value &” in the first quadrant of the unit circle (see Fig. 1.11). A second solution is given by (a — 6), which is in the second quadrant, and is denoted by #°).Displacement analysis of two-link serial manipulators — 17 0% =q-Q0 2nd eo gle>o \ Se] oe 0% = 90-_ gle <0 Figure 1.11 Solutions of sin 6 = g/e. When e/g < 0 a calculator will display a value — 6 in the fourth quadrant. A second solution is given by —(a + (—@)), which is in the third quadrant and is denoted by 6°). Finally, we need to solve the following pair of simultaneous trigonomet- tical equations: Acos 6— Bsin 6= E, (1.6) and Bcos 0+ Asin 0= F, (1.7) where A, B, E, and F are known coefficients. Subtracting B X (1.6) from A X (1.7) and then adding A X (1.6) to B X (1.7) gives, respectively, (AF — BE) sin 0 = (AP +B) * (1.8)Displacement analysis of planar robot manipulators 18 O=n+0, ® 6-9, (cos, . sin®e) = (+) (cos, sin ®.) = (+,+) O=n+0, 7 } @= 20-0 (cos, . sin®e ) = (-r) (cos®, , sin, ) = (+) Figure 1.12 Computation of 6. and (AE + BF) cos 0= (Az +B)” (1.9) provided that either A or B is different from zero. The pair of values (sin @, cos 6) yields a single value for 6. This is easy to compute from the value 6, = tan“! {(AF — BE)(AE + BF)}, whichis ob- tained by dividing (1.8) by (1.9). A calculator will display either a positive or negative value for 6.. It remains to determine in which quadrant the an- gle @ lies. A calculator will display a positive value for 8, when @ lies in the first and third quadrants (see Fig. 1.12), and a negative value for 6 when 0 lies in the second and fourth quadrants. The quadrant can be identified by the ordered pair of the signs of (cos 6, sin 0,), as illustrated by the unit cir- cles in Figure 1.12.Displacement analysis of two-link serial manipulators 19 1.4.1 Analysis of the RR manipulator Throughout, the coordinates of point Q will be denoted by (x, y), which by parallel projection (see Fig. 1.9) on the X and Y axes are X = ayes + 493 C142, (1.10) Y = 41981 + 473 S142. (LA), Also, y= 6+. (1.12) The abbreviations c, = cos 6}, 51 = sin 0), C1+2 = cos (0; + 4), and sj42 = sin (0, + 62) have been introduced. When the joint displacements 6,, @ are known, (1.10) and (1.11) deter- mine a unique location (position and orientation) for the gripper {x, y; y}. This is called the forward analysis.. The reverse analysis is more difficult. We need to specify a position of a tool point Q in the end effector and to compute the joint displacements 6, and 62. This is accomplished by adding the sums of the squares of (1.10) and (1.11), which yields (2 +9?) = ah, + ay + 2ayoars (cr42e1 + 514251). (1.13) Then introduce the identity cy42c + 51425) = Cj+2-1 = co and regroup terms to give dcosh=f, (1.14) where. d = 2a)2493, faety-ay,—a,. (1.15) Equation 1.14 yields two solutions, 6) and —@, or equivalently 6{4-») (see the solution of 1.14), which define a pair of manipulator configurations (see Fig. 1.13). Corresponding pairs of values 6(*») can be computed by solving (1.10) and (1.11) for s; and c). This is accomplished by expanding the right sides of (1.10) and (1.11), and re-grouping terms, which yields (iz + G23¢2)c1 — (42352)81 = x, (1.16) (a2352)¢€1 + (ai2 + a23¢2)1 = Y. d.17)Displacement analysis of planar robot manipulators 20 Figure 1.13 Two RR manipulator configurations. These equations are the same form as (1.6) and (1.7) with A = a}2 + a93¢2, B = ay35, E = x, and F = y. A pair of values 0% which corresponds to 6%) can be computed. By using (1.12), the sets (6{%, 6%) and (6\”, 6”) determine uniquely the orientations +f and ¥°). 1.4.2 Analysis of the RP manipulator By parallel projection on the X and Y axes (see Fig. 1.9), the coor- dinates of Q are x= Sic, + @3 co F + a) = Sic, — 4351, (1.18) y = Sis, + a3 sn($ + a) = Sys + ay3¢}. (1.19) = 6 + wl. (1.20) When the joint displacements 0); S; are known, (1.18}(1.20) determine a unique location for the gripper {x, y; y}. The reverse analysis is performed by adding the sums of the squares of (1.18) and (1.19) and solving for Se which yields =axrty-a, (1.21) Equation 1.21 gives a pair of values, S, = +(x? + y? — @},)!”, which will be denoted by se, Equations 1.18 and 1.19 are the same form as (1.6) andDisplacement analysis of two-link serial manipulators = 21 Figure 1.14 Two AP manipulator configurations. (1.7), with A = Si, B = ay, E = x, and F = y. A pair of values 0{%”, which corresponds to sia), can thus be computed. Finally, a pair of values 7) can be computed from (1.20) (see Fig. 1.14). 1.4.3 Analysis of the PR manipulator By parallel projection on the X and Y axes (see Fig. 1.9), the coor- dinates of Q are x = Sy + a34 cos(m/2 + 63) = S, — a3483, (1.22) Y = G23 + a3q sin(m/2 + 63) = a3 + asqc3, (1.23) y= 03+ m2. (1.24) When the joint displacements { 5); 63} are known, (1.22)-(1.24) determine a unique location for the gripper {x, y; y}. A reverse analysis is performed by solving (1.23) for @3, which yields two solutions, +63 and — 63, or equivalently, 6” (see the solution of (1.4), which defines a pair of manipulator configurations (see Fig. 1.15). Corre- sponding values for the slider displacement S(*” can be computed from (1.22) and a corresponding pair of values y*) can be computed from (1.24). 1.4.4 Analysis of the PP manipulator By parallel projection on the X and Y axes (see Fig. 1.9), the coor- dinates of Q are x= S, + a4, (1.25)Displacement analysis of planar robot manipulators 22 Figure 1.15 Two PR manipulator configurations. and y= S2 (1.26) Furthermore, y=0. (1.27) The forward solution is given by (1.25) and (1.26), whereas the reverse so- lution is given by S) = x — a3q and S, = y. There is a unique solution for a specified value of the coordinates of Q. Complete Exercises 1.1 and 1.2 as far as possible without using the text by employing the following procedure. 1. Draw the skeletal form of the manipulator to be analyzed using the sequence of joints specified, beginning with the first grounded joint. 2. Label the skeletal form with the manipulator dimensions and vari- ables. 3. Employ parallel projection, write expressions for the coordinates (x, y) of a tool in the end effector, and obtain an expression for the orientation of the end effector measured relative to the X axis.Analysis of 3R and 2R-P three-link manipulators 23 4, Perform a reverse analysis by solving the expressions for (x, y) for one joint variable. 5. Compute a pair of sets of joint variables which define the two dis- tinct closures for the end-effector location by employing, where nec- essary, the solutions of the trigonometrical equations 1.4-1.7. EXERCISE 1.1 ‘We want to position the tool points of planar RR, RP, and PR ma- nipulators (see Figs. 1.13-1.15). Manipulator type Dimensions* Coordinates of @ RR az = 1.5, 423 = 0.5 (0.317, 1.549) RP a@3=1 (0.134, 2.232) PR 23 = 2, G34 = 1.5 (—2.207, 1.293) Perform a reverse analysis for each manipulator, compute values for y, and draw to scale in inches the two configurations for each manipulator. Verify your re- sults by performing a forward analysis. 1.5 Analysis of 3R and 2R-P three-link serial manipulators The notation for two-link serial manipulators (see Section 1.4) is now adapted to the three-link RRR, RRP, RPR, and PRR manipulators (see Fig. 1.16).** Some readers may ask why a PPP manipulator is not included in this list. The answer is that a nonrotating PPP assemblage can only produce pure translational motion of the end effector, which can be obtained by ac- tuating a PP manipulator (see Fig. 1.17). A third P joint is not linearly inde- pendent of the first two P joints. * The dimensions and coordinates can be chosen in any consistent units of length. ** Section 1.4 examined cases where the end effector only has two degrees of free- dom. In other words, we are not able to freely choose all three parameters (x, y; y). With the addition of a third independent joint (a P or R joint), it is possible to control the three freedoms of the end effector, specifically by controlling the three joints simultaneously. This section develops the analytics necessary to do this.Displacement analysis of planar robot manipulators 24 RRR / \ RPR \ PRR Figure 1.16 3R and 2R-P manipulators. 1.5.1 Analysis of the RRR manipulator By parallel projection on the X and Y axes, the coordinates of point Q are (see Fig. 1.18) X= ay2 C1 + G93 C142 + G34 C142+3, (1.28) Y = Qy2 81 + G23 S142 F 34 $14243- (1.29) Furthermore, the orientation angle is y= 4+ & + 6. (1.30)Analysis of 3R and 2R-P three-link manipulators 25 Y ¥ 4 4 ee EO g eee Y 5; 52 *S;siny o S, o 5) *S; cos: ae i 2 + 1°53 COSY Ite i} x Ty Figure 1.17 Equivalent PPP and PP manipulators. Figure 1.18 An RRR manipulator. When the joint displacements 6), 6, and 63 are known, (1.28)-(1.30) deter- mine a unique location for the gripper {x, y; y}. A reverse analysis for which {x, y; y} is specified is accomplished by substituting (1.30) into (1.28) and (1.29) and re-arranging in the form X ~ G34 COS Y = ay2C1 + G23C1 +2 (1.31)Displacement analysis of planar robot manipulators 26 Y — 434 sin -y = 4125) + ap351+2. (1.32) Add the sums of the squares of (1.31) and (1.32) to yield (x — ayq cos y)? + (y — agg sin ¥)? = aby + ay + Zarrarx(c1+201 + $14281)- (1.33) Then, we introduce the identity c)+2¢, + 51+25) = C1+2-1 = C2 and regroup the terms, which gives d cos & =f, (1.34) where d = 2aj2423, f= (x — a34 cos y)? + (y — ag sin y)? — at, — a5. (1.35) Equation 1.34 can be obtained directly by applying the cosine law to trian- gle 123 (see Fig. 1.18). Also, (1.34) yields two solutions +6) and —6;, or equivalently, 6{” (see the solution of (1.4)). These two solutions define a pair of manipulator configurations (see Fig. 1.18) which can be constructed using Figure 1.13.* The corresponding values for 6“) can be computed by solving (1.31) and (1.32) for s; and c). If we expand the right sides of (1.31) and (1.32) and regroup by terms, this yields Ac; — Bs, = E, (1.36) Be, + As; = F, (1.37) where A= ay + ap3¢2, E= x — a34 Cos y, B= a35, F=y-— aja sin y. (1.38) These equations have the same form as (1.6) and (1.7). A pair of values 6%” * Note that for the 3R manipulator all three coordinates {x, y; y} are computed and therefore the two configurations are determined with link a34 in a fixed position. It fol- lows that the two configurations can be obtained using the 2R manipulator constructions illustrated in Fig. 1.13. Similarly, other three-link manipulator configurations can be re- lated to corresponding two-link manipulators.Analysis of 3R and 2R-P three-link manipulators 27 Figure 1.19 An RRP manipulator. can be computed, which corresponds to 6”. Equation 1.30 can be used to compute 6°”), which corresponds to sets 6°, 4”, and 6, 6. 1.5.2 Analysis of the RRP manipulator By parallel projection on the X and ¥ axes, the coordinates of point Q are (see Fig. 1.19) X = ay2C) + 493 C142 + S3 cos (m/2 + 8, + &), (1.39) Y = ay251 + 23 5142 + S3 sin (w/2 + A + 02). (1.40) The orientation angle is y= TI2 + 0 + (41) When the joint displacements 6), 6), and S3 are known, (1.39)-(1.40) deter- mine a unique location for the gripper {x, y: y}. A reverse analysis for which Q{x, y; y} is specified is now performed. From (1.41), —s142 = cos y and c)+42 = sin y. Substitution into (1.39) and (1.40) yields xX — a3 Sin y = aj2C, + S3 cos ¥, (1.42) y + a3 cos y = ais; + Sy sin y. (1.43)Displacement analysis of planar robot manipulators 28 la, oO Figure 1.20 An RAP manipulator. Subtract cos y times (1.43) from sin y times (1.42) to get ap sin(y — 61) = x sin y — y cos y — a3. (1.44) From (1.41), sin (y — 01) = sin (6, + 7/2) = cp. Therefore, (1.44) can be ex- pressed in the form do=f, (1.45) where d = aj2 and f= x sin y — y cos y — a3. This yields two solutions, +6, and —6, or equivalently, #6” (see the solution of (1.4)). The corre- sponding pair of values 6@” is computed from (1.41), and a corresponding pair of values S‘*” is computed from either (1.42) or (1.43), and eS in ¥ — arc? = aps X — a3 sin Ay2€) + 93 cos Q28) seh) a 23 SIN _'Y ~ 412% yt a3 Y= 412! . (1.46) cos y sin y Figure 1.20 illustrates another form of the RRP manipulator. It is left to the reader to perform a forward and reverse analysis. 1.5.3 Analysis of the RPR manipulator By parallel projection on the X and Y axes, the coordinates of point Q are (see Fig. 1.21) x = Sic — a35) + a34 cos(7/2 + 8) + 43), (1.47) y = Sysy + ay3¢1 + aaa sin(a/2 + 6; + 63). (1.48)Analysis of 3R and 2R-P three-link manipulators 29 Y 2 6 8+ % O34 = Sse a3 9, Ss; na Figure 1.21 An RPR manipulator. The orientation angle is y= WI2+ 0 + 0. (1.49) When the joint displacements 6,, S;, and 43 are known, (1.47)(1.49) deter- mine a unique location for the gripper {x, y; y}- A reverse analysis is performed using (1.47) and (1.48), which can be re- arranged as Sc, — 4935, = x — a34 COS Y (1.50) Sys, + an3c1 = y — a34 sin y. (51) Adding the sums of the squares of (1.50) and (1.51) yields Si = (x — a34 cos YP + (y — 434 sin )? — a2,. (1.52) Equation 1.52 gives a pair of values, S; = +{(x — a34 cos )? + (y — a3q sin ‘y)? — @23?}!7, which will be denoted by 5”, Equations 1.50 and 1.51 have the same form as (1.6) and (1.7) with A = S‘*) B = ay3, E= x — ay cos y, and F = y — a34 sin y. Thus, a pair of values 6”, which corresponds to S\*»), can be computed. Finally, a pair of values 64” can be computed from (1.49). Figure 1.22 illustrates another form of the RPR manipulator. It is left to the reader to perform a forward and reverse analysis.Displacement analysis of planar robot manipulators 30 — x x oO Figure 1.22 Another form of the RPR manipulator. 1.5.4 Analysis of the PRR manipulator By parallel projection on the X and Y axes, the coordinates of point Q are (see Fig. 1.23) X= Sy + Ay3¢2 + A34C243, (1.53) Y = Gp352 + 034 5243. (1.54) The orientation angle is y= + 0. (1.55) When the joint displacements are known, (1.53)-(1.55) determine a unique location for the gripper {x, y; y}. A reverse analysis is performed by substituting (1.55) into (1.54), which gives 47352 = y — 434 sin y. (1.56) This is in the same form as (1.5), with €=a73, g=y— ax sin y. (1.57) Hence, the solution yields pairs of values of @, which we denote by ae)Analysis of 3R and 2R-P three-link manipulators 31 ca oO Figure 1.23 A PRR manipulator. (see the solution of (1.5)). These define two distinct configurations, which are easy to construct (using Fig. 1.15). The corresponding values 6%” can be computed from (1.55), and the corresponding values for S; can be com- puted from (1.53): Si) = x — aac) — arg COB Y. (1.58) EXERCISE 1.2 1. Locate (position tool point Q and orientate (y) link a34) the end effectors of the RRR, RRP, RPR, and PRR manipulators. Manipulator Dimensions Coordinates Orientation type (ins.) of Q (ins.) ¥y deg. RRR ay = 3,43= 2, (-0.84,4.02) 105 ayy = 0.5 RRP a2 = 3,a33=2 (—3.5, 2.10) 270 RPR ay =2,ayg=1 (2.5, -3.08) 145 PRR ayy = 2, a34 = (-175, 2.7) 105Displacement analysis of planar robot manipulators 32 = \ \ de « © RPP = \ S>S=_— SH PRP PPR = | Figure 1.24 The 2P-R manipulators. Perform a reverse analysis for each manipulator and draw the two configurations. Verify your results by performing forward analysis. 1.6 Analysis of 2P-R three-link serial manipulators The forward and reverse analyses of the three-link RPP, PRP, and PPR manipulators (see Fig. 1.24) are extremely simple. In any one case, there is only a single manipulator configuration for each gripper location {.x, y; y}, because there is a one-to-one relationship between the single revolute joint displacement and y. It is left to the reader to perform these analyses.Analysis of redundant-serial manipulators 33 Figure 1.25 A 4R manipulator. 1.7. Analysis of redundant-serial manipulators with four or more kinematic pairs The example in Figure 1.25 illustrates a planar 4R redundant manipu- lator. By parallel projection on the X and ¥ axes, the coordinates of point Q are X = AyaCy + 33C1+2 + a34C14243 + 2451424344 (1.59) Y = 4251 + Gz38142 + 434514243 + 44581424344 (1.60) The orientation angle is Y= A+ + 03+ % (1.61) When the joint displacements are known, (1.59)-(1.61) yield a unique loca- tion for the gripper {x, y; y}. However, there is an infinite number of solutions for the reverse analysis. The substitution of (1.61) in (1.59) and (1.60) yields = ayacy + €23€1+2 + 434514243 + G45 COS Y, (1.62) and Y = ay251 + 238142 + 434514243 + Gas sin y. (1.63) It is not possible, in general, to obtain a unique solution for this pair of equa-Displacement analysis of planar robot manipulators 34 tions when the gripper location Q(x, y; y) is specified. The 4R manipulator possesses an additional freedom, or more than the three freedoms that are sufficient to locate the gripper. Such manipulators are redundant. The analysis of redundant manipulators has been the subject of much re- search. Briefly, such manipulators should permit sophisticated articulation and motion planning where the extra freedoms can be used to avoid obsta- cles. Proposed solutions to the problem usually include optimization in one form or another. Optimization formulations must be geometrically meaning- ful and thus be invariant under the different selections of reference points, such as O and Q, the orientations of coordinate systems, and they must pro- vide the same results using different systems of units (feet or meters, for ex- ample). An interesting problem occurs when a gripper is performing a cyclic motion, ie., it is repeating a sequence of locations throughout the cycle. It is desirable for any optimization formulation to determine manipulator config- urations throughout a cycle which are reproduced in subsequent cycles. A meaningful solution to redundant serial manipulators that accomplishes this is given by Chung, Griffis, and Duffy (1994). 1.8 Displacement analysis of in-parallel manipulators 1.8.1 Description of in-parallel manipulators The most general form of an in-parallel planar manipulator with three parallel, serially connected RPR chains joining a fixed base to a movable lamina is shown in Figure 1.26. The term in-parallel means that each serial chain has the same structure and one joint is actuated in each chain (for in- stance, the prismatic pair). In contrast to serial devices it is relatively simple to perform a reverse analy- sis, i.e., it is simple to compute a unique set of connector lengths 1 — 1’ = £1, 2-2' = 5, and 3 — 3' = €3 for a specified location of the moving lamina. However, the forward analysis of the general in-parallel manipulator is alge- braically complicated, and we need to compute the multiple locations of the moving lamina for a specified set of connector lengths €;, €2, and €3. The so- lution can be expressed in the form of a sixth-degree polynomial, which means that there can be up to a maximum of six different real assembly configura- tions of the mechanism for a specified set of connector lengths. Such an analy- sis is beyond the scope of this book, However, a method for deriving a sixth-Displacement analysis of in-parallel manipulators 35 Figure 1.26 A general in-parallel manipulator. z > r 2 Figure 1.27 A model for a three-finger grasp. degree polynomial is given in Li and Matthew (1987). This mechanism can be used to model a three-fingered grasp (see Fig. 1.27). The central movable triangle is formed by the contact points of three, two d.o.f. RR fingers. Each contact point is also modeled by a revolute (R) kinematic pair. Two in-parallel manipulators with special geometries are illustrated in Fig- ures 1.28 and 1.29. The gross motions or work spaces of in-parallel manipulators are smallerDisplacement analysis of planar robot manipulators 36 3 2 1 r 2 x Figure 1.28 An in-parallel manipulator with special geometry (first example). 2,3 Hf 1,2 ia Figure 1.29 An in-parallel manipulator with special geometry (sec- ond example). than those for serial manipulators. However, parallel devices can sustain much greater forces than serial devices. Generally, the advantages of in-parallel ma- nipulators are that they are more accurate in positioning and orientating work- pieces than serial manipulators, they possess a high payload/weight ratio, and they are easily adaptable to force and position control. 1.8.2 A reverse analysis for the general in-parallel manipulator As we mentioned earlier, it is relatively simple to perform a reverse analysis for this device. The location {x, y; y} of the movable lamina is specified, where (x, y) are the coordinates of a point Q on the movable plat- form and y is the orientation of the platform, measured relative to the X axis.Displacement analysis of in-parallel manipulators 37 815 Ox. y) au SF 3 (X,Y) (X35, 0) o Be Figure 1.30 A line diagram for a general in-parallel manipulator. Here y is measured using side 13. Since point Q is fixed on the moving plat- form, then g; and £ (see Fig. 1.30) are known as well as the parameters 6, 812, and gi3 which specify the triangular movable platform. What follows is one of many possible procedures for determining the connector lengths €1, 2, and €3, It is immediately apparent from Figure 1.30 that x1 =X— g1 Cos (B+ 9), d= ym gisin(B + y). Also, X2 = x1 + gy cos (6 — ), J2= Yi 812 sin (S— y), and x3 = x1 + gi3 cos y, Ys = Ya + 813 sin y. (1.64) Finally, €= G3 +99", = (i — x? +02—y,P)%, and 6 = {os - xi)? +95).Displacement analysis of planar robot manipulators 38 1,2! ay 3 a2 Figure 1.32 Two assembly configurations above the base ajp. 1.8.3 A forward analysis for an in-parallel manipulator with the simplest geometry As we mentioned before, it is complicated to perform a forward analysis for the general in-parallel manipulator. However, the analysis is relatively sim- ple for this manipulator with special geometry (see Fig. 1.31). For a speci- fied set of connector lengths £1, €2, €3 and sides ay2, a},, two distinct as- sembly configurations above the base can be obtained. Point 1” is a reflectionDisplacement analysis of in-parallel manipulators 39 Figure 1.33 Two assembly configurations below the base ai. of point 1 through the connector €2 (see Fig. 1.32). There are two additional configurations (see Fig. 1.33) that are reflections through the base. The four assemblages in Figures 1.32 and 1.33 can be computed using the following pair of cosine laws (see Fig. 1.31): cos (0 — O) = (€2 + €3 — at,V/2€ rls, cos @ = (£3 + a't, — €3)/2bpa1». (1.65) All four assembly configurations can be drawn if the angles @, and @, are known. T=. 4 1; 15-4 te eearereraanaaronroroperse rsa as ano op eSreTSra aT SISTSOTSPETECEEERT ETSI Solve (1.65) and obtain all four assembly configurations of the simplest in-parallel manipulator with a}, = 3, a12 = 1.35, €1 = 1.95, €2 = 3, and £3 = 2.2 (ins.). Verify these results by construction using only a ruler and a compass.2 Planar statics 2.1 The coordinates of a line in the XY plane Assume that the page is the XY plane and the Z axis is pointing upward out of the page through Q. Most students are familiar with using point coordinates, for example, how to designate points 1 and 2 by the coordinates (x1, y1) and (x2, y2) (see Fig. 2.1). The coordinates of a point are essentially an ordered pair of real numbers. This may be something new for some students: we will determine the coordinates for designat- ing a line. The two vectors rj and rz from O to points 1 and 2 with coordinates (x1, y1) and (x2, y2) determine the directed line segment 12 with vector S given by S=n-m (2.1) The projections of S onto the X and ¥ axes are, respectively, L=x.~x,=(|S|cos@ and M=y2—y, =|S| sin 6, (2.2) where |S) =(2 + my. (23) LS| and MI|S| are the direction cosines of the line segment. When |S| = 1, the directed line segment has unit length L? + M? = 1, and then L = cos 6 and M = sin @. 40Coordinates of a line in the XY plane 41 Figure 2.1 A line in the XY plane. The moment of the directed line segment 12 about the origin is given by the vector product r X S, where r is any vector drawn from O to any point on the straight line joining points 1 and 2, and the vector S can be located anywhere on the line. Now from Figure 2.2, m1 XS = |r] - [S| sin 1 k = ([m] sin $1) [Sik = |p| [Slk, Fo X S = |r| [S| sin go k = (|r| sin 2) |S|k = |p| [SI XS = |r| - [S| sin ok = ({r] sin 4) [S[k = |p| [Sik and Pp XS = |p| - |S| sin 7/2 k = |S|k = |p| [Slk, where k is a unit vector parallel to the Z axis. Therefore, the moment can be expressed in the alternative forms rxS=r,XS=mxXS. (2.4)Planar statics 42 Figure 2.2 Moment of directed line segment about O. The left and intermediate sides of (2.4) can be expressed in the determinant forms ijk ij k| |i i k x y O;= |x yO} = ba y 0 LM O LM O 2-1 YM O ijk =| yn Of], (2.5) x2 y» 0 where j and i are unit vectors parallel to the X and Y axes, and x and y are the components of r. If we equate the left and right sides of (2.5), the i and j components van- ish identically, while the k component yields the equation for the line, which can be written in the form Ly — Mx+R=0, (2.6) where m1 ee : 27 x2 2 oo It is left to the reader to deduce (2.6) using the left and right sides of (2.4).Coordinates of a line in the XY plane 43 In addition, expanding the right side of (2.5) and equating to the left side of (2.4) yields rXS=Rk. (2.8) The moment vector of the line segment about the origin is parallel to the Z axis and is thus perpendicular to the XY plane. For convenience, the vector Rk in Figure 2.2 is associated with the Z axis. However, subsection 2.4.1 demonstrates that Rk is a free vector which can be associated with any line drawn parallel to that Z axis.* The three numbers L, M, and R were first established by Pliicker, so they are called the Pliicker line coordinates. They are homogeneous because sub- stituting AL, AM, and AR, where A is a nonzero scalar into (2.6) yields the same line. However, their units are not consistent. L and M have dimensions of (length)!, while R has dimensions of area, (length)”. Because of this lack of consistency in dimensions, the coordinates are represented by the ordered triple of real numbers (LZ, M; R) with the semicolon separating R from L and M. It is important to recognize that if the quantities L, M, and R are known, there remains one degree of freedom in the location of the directed line segment 12, because it takes four magnitudes (x, y;) and (x2, yz) to fix the points 1 and 2. The same triple (L, M; R) is obtained if and only if the vector (rz — r}) is free to move on a definite straight line, and this line is determined when only the ratios L: M: R are known. It should now be clear that the only changes of the position vectors r; and r2 that leave (L, M; R} unchanged are translations of the line segment along its line which preserve its length and sense. Such a line segment, which is deter- mined by the ordered triple of real numbers {L, M; R}, is called a line bound vector. The two coordinates (L, M) by themselves determine a free vector, be- cause they are unaltered by a parallel translation of the line segment outside the line. The ratios L; M: R (see also (2.6)), which are equivalent to two quan- tities, determine the straight line and not the length of the segment upon it. * The symbol \) is intended to convey the sense of the turning moment R about k and not the moment vector Rk itself, which is a vector pointing outward and normal to the XY plane (for R > 0).Planar statics 44 Figure 2.3 Triangle in the XY plane. The line bound vector {L, M; R} was elegantly represented by Grassmann by the three 2 X 2 determinants of the matrix lua yw 1 x2 yf obtained by deleting, in turn, the columns 1] ]) ona ae day Le. 1 _ fl om a=|! Furthermore, the area of a triangle with vertices 1, 2, and 3 and coordinates (x1, 1), (2, y2), and (x3, y3) (See Fig. 2.3) is given by Thus lox 1 x =f x2 2 =| Tou yi 1m ye 1 x3 y3 A= : (2.9) vie Therefore, the area of the triangle O12 (see Fig. 2.1) is given by oes 2 M1 1 M2 2 eae 7k (2.10)Coordinates of a line in the XY plane 45 Figure 2.4 Determination of p. Therefore, R is twice the area of the triangle O12 taken in the sense O, 1, 2. From Figure 2.1: A=Spbl=pp@+ay, (2.11) where p is the length of the vector p drawn from O perpendicular to the line. Comparing (2.10) and (2.11) yields greene eat P+ my Therefore, for a unit line segment for which |S| = (L? + M?)¥2 = 1 we have R = p and the line coordinates are {c, s; p}, with the abbreviations c = cos @ and s = sin 6. If we substitute these results into (2.6), the equation for the line can be expressed in the form. (2.12) cy — sxt+p=0. (2.13) If the coordinates of any point (xg, yg) or (xg, 0) on the line are known together with 6 (see Fig. 2.4), the value for p is easy to obtain from (2.13)Planar statics 46 S=(c.s') Figure 2.5 Determination of the line coordinates. and p = sxg — cyg = sxg. These expressions for p are easy to verify by sim- ple projection using Figure 2.4. The coordinates for a line can be measured in two distinct ways. The same line $ is illustrated in Figure 2.Sa and b. The homogeneous coordinates for the line using Figure 2.5a are {cos 6, sin 0; p}, where p = xg sin 8 — yg cos 0, whereas the homogeneous coor- dinates for the line using part b are {cos 6’, sin 6’; p'}. Now, 6’ = 6+ 7 and, hence, cos @’ = —cos 6, sin 6’ = —sin@, and p’ = (xg sin 6’ ~ yg cosCoordinates of a line in the XY plane 47 Y IN A = 1) o UN - Figure 2.6 Generation of a pencil through the point with coordinates 0. 2 Pins 1 pa ee i L 2 8 rads 0.00 157 3.14 4.71 6.28 Figure 2.7 Plot of p versus 6. 6') = —(xg sin 6 — yg cos ) = —p. Thus, the homogeneous coordinates for the line $ using part b of Figure 2.5 are {—cos 6, —sin 6; —p}, which are the same as {cos 6, sin 6, p}. It is, however, important to recognize that p can assume positive or negative values. Assigning a sign to p is easily done by inspection. In part a of Figure 2.5 the unit vector S is turning about O in an anticlockwise sense and so p is a positive quantity, whereas in part b the unit vector S' is turning about O in a clockwise sense and, hence, p is a negative quantity. It is interesting to plot p versus @, and this is shown in Figure 2.6 for the pencil that passes through the point Q with coordinates (1,1) in Figure 2.7.Planar statics 48 Consider that the pencil is formed by rotating the line labeled $ about Q with the vector S attached to it. 2.2 The coordinates for the point of intersection of a pair of lines Consider the pair of lines Liy — Mix + Ri = 0 (2.14) and Toy — Mox + Rp = 0. (2.15) Then eliminate y by subtracting L; times (2.15) from Ly times (2.14), which yields (LiM2 — LnMi)x + (Rilo — Roly) = 0, or (LiM2 — 1pM))x = (EiR2 — L2R))- (2.16) Eliminate x by subtracting M, times (2.15) from M3 times (2.14), which yields (LiMz — LnMy)y + (RiM2 — RoMy) = 0, or (LiMz — LpMy)y = (MyR2 — MR:). (2.17) It follows from (2.16) and (2.17) that the coordinates for the point of inter- section can be expressed by My, Ry, Mr Ro Ly; Ril |2i Mi Ln Ro| |Lln M2 The sequence of determinants on the right side of (2.18) can be obtained di- rectly from the matrix of line coordinates using Grassmann’s expansion of the 2 X 2 determinants of Ly M, Ri I, Mm, Rl by deleting, in turn, the columns [bli == [e} yix:l= ; (2.18)Statics of plane rigid systems 49 The ratios of the coordinates of the point of intersection of a pair of lines with normalized coordinates {c,, 51; pi} and {c2, 82; pz} can be obtained di- rectly from the matrix cst PL co sz poy Si Pa S2 P2| and cr Pi Jeo pal” ey Sy cz, $2, : (2.19) EXERCISE 2.1 1. On separate figures draw the lines that join the pairs of points (1,1), (—2,—2); (4,1), (—2,—3); (— 1,3), (3,2); and (—2,2), (1,-5). Using Grassmann’s expansion of the 2 X 3 matrix Ton y [ie 3] determine the two sets of Plticker coordinates {L, M; R} for each line obtained by interchanging the rows of the matrix. Label each line with the pair of equa- tions Ly — Mx + R = O and compare your results. In each figure draw the tri- angle O, 1, 2 and determine the signed area O12 using (2.10). Compare these values with the corresponding values of R you have obtained. Express each set of Pliicker coordinates in the unitized form {c, s; p}, and write the pair of equa- tions for each line. Label each line with the directed angle 6 and the directed per- pendicular distance p. Using Grassmann’s expansion of the 2 X 3 matrix LM RX I, Mr Rr determine the coordinates of the intersections of the pairs of lines (3.03, 1.75; 3.5}, (—1.03, 2.82; —6}; and {—2.82, 1.03; —2.1}, {—1, 1.73; 3.4}. Verify your results by drawing each pair of lines. 2.3 The statics of plane rigid systems The concepts developed in the previous section can now be applied directly to the statics of planar rigid systems. The directed line segment S can be considered equivalent to a force applied to a rigid lamina (see Fig.Planar statics 50 Figure 2.8 Representation of a force on a rigid lamina. 2.8). Because the lamina is rigid, the point of application can be moved any- where along the line. The first two coordinates of the directed line segment {L, M; R} are called the components of the force, and the length of the line segment (L? + M*)"? is the magnitude of the force. From (2.12) the turning moment R is the product of the distance p and the magnitude of the force. The problem of determining the resultant of an arbitrary system of forces {L;, My Ri}, i= 1, 2,..-n, that acts upon a plane lamina is essentially that of determining a unique line bound vector with the coordinates L=S, M=SM, R=DR. (2.20) 1 1 =i However, there is an important exception. Consider the resultant of a pair of equal and opposite forces with the coordinates (L, M; Ri} and {—L, —M;, Ro}, where R, # —Rp (see Fig. 2.9). The coordinates of the resultant (0, 0; R, + Rp} are not a line bound vector, but a pure couple. Clearly, (L? + M?)!? = 0 and from (2.12), p = ©. As illustrated by Figure 2.9, a couple can be considered a force of infinitesimal magnitude, |f| = (L? + M?)? > 0, acting along a line which is parallel to the lines of action of the pair of forces and is infinitely distant, p = ©, such that |8f| p = (R, + Ro). This line is calledStatics of plane rigid systems 51 Pxdf =(R/-Ro)k Figure 2.9 A pure couple. the line at infinity,* and the coordinates of the resultant couple {0, 0; Ry + R2} can be expressed as (Ri + R2) (0, 0; 1}, where (Ri + R2) is the magni- tude of the resultant and {0, 0; 1} are the coordinates of the line at infinity in the XY plane. A pure couple can thus be represented as a scalar multiple of the line of infinity, and hence there is no exception to the addition of forces in the XY plane. This representation is completely compatible with the fact that p X df = (R; + Ro) k, which is a free vector associated with the direc- tion k and not with a definite line. It is of interest to pursue this example by considering a pair of antiparal- lel forces that act upon a rigid lamina (i.e., a pair of forces which acts on par- allel lines but in the opposite sense). We assume, without loss of generality, that the lines of action of the forces are parallel to the Y axis (see Fig. 2.10) and their coordinates are thus given by {0, My; Ri} and (0, —Mz; —Ry}, or M,(0, 1; py} and —M2{0, 1, po}, where M; # 0, and Mp # 0. The magnitude of the resultant is given by M=M,- M2. (2.21) * For more details on the line at infinity, see Klein (1939) and Hunt (1990), Briefly, in the Euclidean plane a pair of parallel lines do not meet. This is an exception to the gen- eral statement, “the meet of two lines is a point and the join of two points is a line.” In- troducing the line at infinity overcomes this difficulty and pairs of parallel lines meet at points at infinity, all of which lie on the line at infinity.Planar statics 52 y4 Figure 2.10 The resultant of a pair of antiparallel forces. Furthermore, Mp = Myp, — Mopo, (2.22) and therefore the line of action of the resultant can be denoted by its distance Pp from O, M = 1p, -— 2 pr. 2.23) P ™, - M,”" M, - My”? (2.23) The substitution of A = My/Mp in (2.23) yields =p - (2.24) P ea PL ed P2- ro In Figure 2.10 the rigid lamina is represented by a contour drawn on the XY plane, and the lines of action of the applied antiparallel forces pass through points A and B. Let E be the point where the line of action of the resultant meets the X axis. The combined effect of applying the antiparallel forces upon the lamina is thus equivalent to extending the boundary of the lamina and applying a single resultant force through the point E. In this sense, the lam- ina is considered to be unbounded. Assume that M> > M,, and M) increases. The line of action of the resul-Statics of plane rigid systems 53 M,? M, B mai , M,>M, Figure 2.11 Plot of M versus p. tant will approach B. From (2.24), when M, > ©, then A = 0 and the line of action of M becomes p = po, and thus the resultant passes through B. Addi- tionally, when M, = Mp, then A = 1, and from (2.24) p > ©, and the line of action of the resultant lies on the line at infinity. The M versus p plot is the rectangular hyperbola labeled Mz > M,, as illustrated in Figure 2.11. Assume now that M, > M). It is convenient to express (2.24) in the form eae (a) PY T- aay" T- am (2.25) As M, increases, the line of action of the resultant will approach A. From (2.25) when M; — ©, then 1/A = 0 and p = py. The line of action of the re- sultant thus passes through A. Furthermore, when M) = M), A= 1, from (2.25) p — , and the line of action of the resultant lies on the line at infin- ity. The M versus p plot is the rectangular hyperbola labeled M, > Mp, as il- lustrated in Figure 2.11. The resultant force thus lies to the left of point A or to the right side of point B, and it may not physically act upon the lamina. The forces at A and B are thus equivalent to a single force that acts upon a line as if the lamina were extended to include the line of action of this resultant force. Clearly, the resultant could lie on the line at infinity, and in statics it follows that the lamina has no boundary, but rather it can be considered infinitely large.Planar statics 54 Figure 2.12 The resultant of a pair of parallel forces. The magnitude of the resultant of a pair of parallel forces (0, Mi; Ri}, {0, Mz; Ro} or M,{0, 1; pi} and Mp {0, 1; p2} (see Fig. 2.12) is given by M=M,+ M). (2.26) The line of action of the resultant is given by Mp = Mp; + Mop2, and therefore M ~My Ph + My” Ot + My (2.27) It is a simple matter to divide the quotients on the right side of (2.27) above and below by M, and then by M3, and to deduce that as M, — ~, p — p, and as M, > ©, p — pp. Also, when M, = M2, p = 1/2 (p; + p2). The M versus P plot is illustrated by Figure 2.13. It is interesting to note that the resultants of antiparallel and parallel forces lie on pencils of parallel lines. In addition, the resultant of a pair of intersecting forces with coordinates {L1, My; 0}, {L2, Mz; 0) (see Fig. 2.14)Statics of plane rigid systems 55. 0.5(P)*P2) Figure 2.13 Plot of M versus P. {L,+L,, M,*M, + O} {L,, My, O} (Lz, Mz+ O} Figure 2.14 The resultant of a pair of intersecting forces. lies on a pencil of lines that passes through the point of intersection, 0 (see Fig. 2.15). In the three cases illustrated by Figures 2.10, 2.12, and 2.14, it is clear that the resultant is a linear combination of a pair of forces. In other words, any force in a pencil of forces is linearly dependent on any pair of forces in the pencil which does not lie on the same line. It is important to recog- nize that any force which does not belong to a particular pencil cannot be expressed as a linear combination of any number of forces in the pencil. Such a force is linearly independent of the set of forces which constitutes the pencil.Planar statics 56 Figure 2.15 A pencil of lines through point O- 2.4 ~~ Parallel translation and rotation of rectangular coordinate systems Assume that the two points 1 and 2 with coordinates (x, y;) and (22, y2) have the coordinates (x}, y;) and (x3, y3) before and after some trans- formation of the coordinate system. The coordinates of the line bound vec- tor that join points 1 and 2 before and after the transformation are, respec- tively, L=xm-x, M=y.—-y, R= x1y2- 21, (2.28) and L'=xh—x, M'=yy— yy, R= x2 — x291- (2.29) 2.4.1 Parallel translation of a rectangular coordinate system The translation is illustrated by Figure 2.16, and it is clear that m=xy ta m=x, ta, ‘ ; (2.30) N=aN+tB w=yz tb.Rectangular coordinate systems 57 (X25 ¥2)+ (2. Y2) C1, yt), (81, Yi) x Figure 2.16 A parallel translation of a rectangular coordinate sys- tem. It should also be apparent that the points 1 and 2 do not change under this transformation. Therefore, neither does the directed line segment 12 (or force) nor the line joining the points 1 and 2. All of these geometric elements, point, directed line segment or force, and unlimited line, are therefore invariant with the translation of the coordinate system. Substituting (2.30) in (2.29) and comparing with (2.28) yields L=L', M=M, R=R-L'b+Ma, (231) which can be expressed in matrix form as L 1 0 O} LV’ M{=] 0 1 Of{m’}. (2.32) R —b a 1ILR The matrix equation (2.32) expresses the coordinates of the force in the old coordinate system in terms of the coordinates of the force in the new system. Thus far, coordinates have been represented by {L, M; R}. These coordi- nates now appear as the column vector on the left side of (2.32) and are rep- resented by L Mi. RPlanar statics 58 The two coordinates (L, M) remain unchanged under translation and are there- fore independent of the coordinate system. They define a free vector and their values remain unchanged under parallel displacement. As stated earlier, the three coordinates {L, M; R} define a line bound vector. The coordinates {L’, M'; R’} define precisely the identical line bound vector, line segment (or force) in the new coordinate system. Clearly, the value of R changes and is origin dependent. Also, R’ is simply the quantification of the moment of (L, M) about a second origin O'. For a pure couple, L = M = 0 and its coordinates are {0, 0; R}. Let us make this substitution in the latter relation of (2.31) to yield R = R’. A pure couple with coordinates Rk remains unchanged under parallel displacements and is therefore a free vector. We can now give a more direct derivation of the transformation matrix in (2.32). The transformation is of the form L a1 412 413 |} L M|=|an ax ay || mt’ |. (2.33) R 43) 432 a33.LR’ The line coordinates of the X' and Y’ axes in the O' X’ Y’ coordinate sys- tem are represented in matrix form by the column vectors 1 0 O} and | 1 0 0. The moments of the X’ and Y' axes about O are, respectively (see Fig. 2.17), by r x i= —bk and r X i = ak, where r = ai + bi. Therefore, the column vectors that represent the X’ and Y”’ axes in the OXY system are respectively 1 0 0 and | 1 —b. a These coordinates can be obtained directly from an inspection of Figure 2.17: The direction cosines for the X' and Y’ axes are (1,0) and (0,1), re- spectively, while their moments about O are —b and a, respectively.Rectangular coordinate systems 59 Figure 2.17 Line coordinates for the X’ and Y’ axes. The substitution of these results into the left and right sides of (2.33) yields 1 ay a2 a3 J} 1 O |=] 421 a2 an; |] 0 |, (2.34) —b 43, 432 a33IL0 and 0 41 42 a3 |} 0 1} =] 41 22 a3 |] 1 |. (2.35) a 431 432 a334L0 It follows from (2.34) and (2.35) that the first and second columns of the transformation matrix are, respectively,* ay 1 ay, }=] 0 |, (2.36) a3) —b. * There is a well-known result in linear algebra which states that the column vectors of a matrix of a linear transformation represents the image of the basis vectors ih = E in terms of the new basis.Planar statics 60 and a2 ayn} =] 1]. (2.37) 432. a The first two columns of the transformation matrix are the line coordinates for the X' and Y' axes in the old coordinate system. The third column is the line coordinate for the line at infinity, which remains unchanged under any transformation: 0 a1 a2 a13]} 0 O |=] 421 22 a3} ] 0 |, (2.38) 1 43) 432 a334L1 and, hence, a3 0 a3 |=] 0}. (2.39) a33 1 These results yield (2.32) precisely. 2.4.2 Rotation of a rectangular coordinate system Assume that the X’Y' coordinate system is rotated through an angle ¢, as illustrated by Figure 2.18. From the figure: x' =r' cos(at+ d)=r' cos acos P— r' sin asin ¢, (2.40) y =r sin(a+ ¢)=r' sin acos $+ r' cos asin ¢, and x"=r cosa, y'=r' sina. (2.41) The substitution of (2.41) in (2.40) yields the two expressions for the rota- tion x’ =x" cos f— y’ sin g, (2.42) y =x" sin 6 + y" cos ¢. Points 1 and 2 (see Fig. 2.19) do not change under this transformation andRectangular coordinate systems 61 Yr Figure 2.19 Transformation of line coordinates due to a rotation of the coordinate system. neither does the directed line segment 12 (or force) nor the line joining the points 1 and 2. All these geometric elements are therefore invariant with a rotation of coordinate system. The substitution of (2.42) into L’ = x3 — x}, M' = yy — yi, and R’ = x1yy — x4’, yields L' =L' cos ¢— M" sin ¢, M' =L" sin 6 + M’ cos o, (2.43) R=R’,Planar statics 62 which can be expressed in matrix form as L’ cosd sind O]}} L" M'|=|sin@ cosd O|| mM’). (2.44) R 0 ol R" The matrix equation (2.44) expresses the coordinates of the force in the old coordinate system in terms of the coordinates of the force in the new co- ordinate system. The first two relations between (L', M’) and (L", M”) can be obtained directly from Figure 2.19 by parallel projection of L’ and M' on the X" and ¥" axes. For this transformation of coordinates, R remains unchanged. This is not a surprising result because the axis of rotation, the Z axis pointing outward through O, is invariant under this transformation. The two coordinates (L", M") of the vector in the new coordinate system depend solely upon the original val- ues (L’, M’), and the relation of (L’, M’) to (L’, M") is the same as (x', y') to (x", y’) (compare (2.42) and (2.43)). The pairs of values, (L”, M") and (L', M’), quantify the same vector in the two coordinate systems. Finally, it is evident from (2.43) that {L’, M’; R'} are linear homogeneous functions of {L", M1"; R"}, and the ratios L’ : M’ : R’ depend solely upon the ratios L” : M” : R": oe P = cos ¢, — v a ot M” - sin @ + cos 4 =1. (2.45) Ww We have established that these three ratios (without regard to their actual val- ues) determine the line. A more direct derivation of the transformation matrix in (2.44) can now be given. The transformation takes the form Lv 4 42 a3 ]} L" M'|=]an ay || m']. (2.46) R 431 432 331 LR” By analogy with (2.36) and (2.37), the first two columns of the matrix are the line coordinates of the X”, Y" axes in the X’, Y' system, which by in- spection of Figure 2.19 are an cos @ an, | =] sin d J, (2.47) 431 0.Rectangular coordinate systems 63 and a2 cos( + 7/2) —sin b ay, | =] sinh + 7/2) |=] cos o|- (2.48) 32, 0 0. By analogy with (2.39), the third column consists of the line coordinates for the line at infinity, 13 0 ax, |=] 0]. (2.49) 433, 1 2.4.3 The Euclidean group of motions All combinations of linear translations in the plane and rotations around the Z axis constitute the group of Euclidean motions in the plane, which, upon combining (2.30) without the subscripts 1 or 2 and (2.42), can be expressed in the form x=x" cos b— y' sing +a, (2.50) y=x'singd+y"' cos ¢+b, or in matrix form as x cos —sing al][x’ yl=]sin gd cos@ bf] y"}. (2.51) 1 0 ol 1 It is probably self-evident to the reader that scalar geometrical configurations, such as the distance between a pair of points, the angle between a pair of lines, and the area of a triangle, must remain invariant under the Euclidean group of motions. We have shown in previous sections that directed line seg- ments (line bound vectors) which are geometrically equivalent to a force, free vectors such as couples which are geometrically equivalent to area, and the line, all remain invariant with the Euclidean group of motions. We can con- sider such quantities meaningful geometrical configurations in precisely the same way as the scalar quantities: distance, angle, and area. These considerations stem from the two fundamental principles of Grass- mann (see Klein 1939).Planar statics 64 First Grassmann Principle. The geometric properties of any fig- ures must be expressible in formulas which are not changed under a transformation of the coordinate system. Conversely, any formula which, in this sense, is invariant under coordinate transformations must represent a geometric property. A coordinate system is clearly arbitrary (noninvariant in the previous sense) and does not represent a geometric property. Second Grassmann Principle. If the system of magnitudes such as L', M', and R' formed from the transformed coordinates of the points 1 and 2 expresses itself exclusively in terms of the magnitudes L’, M’, and R" formed in the same way from the initial coordinates (the co- ordinates themselves do not appear explicitly), then the system defines a geometric configuration, i.e., one which is independent of the coor- dinate system. In fact, all analytic expressions can be classified ac- cording to their behavior under coordinate transformations, and two expressions which transform in the same way (such as expressions for area and turning moment) are defined as geometrically equivalent. 2.5 Induced force/line transformation under the Euclidean group From (2.32) and (2.44), the coordinates for a force in the old coor- dinate system expressed in terms of the coordinates of the force in the new coordinate system for both a translation and rotation are given by L Lv M|=l[el] M’ |, (2.52) R R" where 1 0 O}Jce -s 0 fel=]| 0 1 0 sc 0 -b ail 0 0 1 c -s 0 = s c Oo}, (2.53) as—be actbs 1 and the abbreviations are c = cos ¢ and s = sin @, as introduced earlier.Induced force/line transformation 65 ye Yr Exercise Figure 2.2(1) Exercise Figure 2.2(2) EXERCISE 2.2 1. Write the Pliicker coordinates for the X’ and Y’ axes in the OXY coordinate system and the Pliicker coordinates for the X and Y axes in the O’X’Y' coordi- nate system (see Exr. Fig. 2.2(1)). Determine the matrices [A] and [A’] for the induced force transformations ole] = bok for pure translations of the coordinate systems. Show that [A’] = [A]~!. 2. Write the Pliicker coordinates for the X" and Y” axes in the OXY’ coordinate system and the Pliicker coordinates for the X', Y’ axes in the OX”Y” coordinate system (see Exr. Fig. 2.2(2)). Determine the matrices [B'] and [B”] for thePlanar statics 66 Figure 2.20 A vector representation of a force. induced force transformations nll ~ Ebel for pure rotations of the coordinate systems. Show that [B"] = [B']“!. 3. Determine [e] = [A] [B']. 2.6 A.useful vector representation of a force We have established that the three coordinates {L, M;, R} represent a force with respect to some reference coordinate system, and the value for Ris dependent on the choice of origin O. It is convenient to represent a force in the vector form with the coordinates (see Fig. 2.20) Ww = (feo). (2.54) The two representations are related by f= (Lit Mj), c. = Rk. (2.55) The subscript 0 is introduced to indicate that the moment vector is origin de- pendent. Furthermore, f can be expressed as a scalar multiple fS of a unit vector S, |S| = 1, where f = |f]. Then the moment vector ¢, = fr X S can beStatics of a parallel manipulator 67 expressed in the form ¢, = fS,, where r X S = S. The pair of vectors (S; S,) thus determines a directed line segment of unit length, and they can be expressed by $= (S; So}. (2.56) The coordinates for a force can now be expressed by a scalar multiple f, the magnitude of the force, and a unit line vector: W = (£ eo} =f (S; So} = fs. (2.57) This latter representation separates the magnitude of the force f, from the geo- metric quantity $. 2.7 ‘The statics of a parallel manipulator The upper lamina is connected to the fixed base via three parallel RPR kinematic chains (see Fig. 2,21). The prismatic pair in each chain is actuated and the platform has three degrees of freedom. Each prismatic pair clearly con- nects a pair of revolute joints, one embedded in the fixed frame and the other in the movable lamina. From here on, this prismatic pair will be called a con- nector, and each force generated in a connector will be called a connector force. We assume at the outset that the geometric configuration is known. Consider that the three connector forces with magnitudes fi, fo, and f; are generated in each of the lines $), $2, and $3 with the coordinates $1, $2, and 53. The upper lamina will experience a force of magnitude f acting on a line $. Therefore, W= wt We. + Ws. (2.58) Equation 2.58 can be expressed in the alternative forms “EEE “s £0 c1 2 3 els - . =fiSi + fsa + Ass £o, _ fs: Sy Ss =A) | +A Ce andPlanar statics 68 Figure 2.21 A parallel manipulator. Therefore, from (2.60): f= fiS; + fpS2 + fiSs, (2.61) and © = fiSo1 + f2802 + fa803. (2.62) From (2.61): F= If] = [fiSi + AS> + ASs|, (2.63) and therefore 5 = F514 So + fSs | (2.64) irl From (2.60) and (2.61): Sy = fiSo + fin + fSo3 (2.65) It is often convenient to express (2.60) in the expanded form 1 C2 ry w=fil 1 | +Al so] +A] 53 |, (2.66) PL P2. P3Statics of a parallel manipulator 69 which can be expressed in the popular matrix form by w=ja, (2.67) where C1 c2 &3 G=| 51 92 53 (2.68) Pi P2 P3 is a 3 x 3 matrix, and fi A=|f (2.69) 3. is a 3 X 1 column vector. When connector forces of magnitudes fj, f:, and f; are generated for some specified configuration, the coordinates w of the resultant acting upon the moving platform can be computed from (2.67). Hence, the magnitude of the resultant f together with its line of action can be determined (see Fig. 2.22). This is called the forward static analysis. Clearly, for equilibrium an exter- nal force of equal and opposite magnitude f must be applied to the platform on the same line. Conversely, when an external force with magnitude f on a line $ is ap- plied to the moving platform we need to determine the magnitudes fi, fo, and fh of the resultant connector forces (see Fig. 2.23), which is easy to obtain from the equation Aap (2.70) where j~! denotes the inverse of j. This is called the reverse or inverse sta- tic analysis. For equilibrium, the equilibrant connector forces are equal and opposite. Assume now that a connector is an RPR kinematic chain joining a fixed and a movable platform at B; and C; (see Fig. 2.24) and that an internal force can be generated by a hydraulic system in the sliding joint, Furthermore, a force f; is generated in the cylinder which acts upon the piston at point £;. For equilib- rium, an equal and opposite reaction force is transmitted from the movable plat- form to the connector at point C;. The connector link C;£; is either in compres- sion (see Fig. 2.24a) or in tension (see Fig. 2.24b), Analogously, the generatedPlanar statics 70 (applied) f Fea sultan (applied) f , $2 Figure 2.22 Forward analysis. ; £ (applied) f, (resultant) Figure 2.23 Reverse analysis. force f; acts upon the cylinder head at point H;, and an equal and opposite re- action force is transmitted from the fixed platform at point B; to the connector BiH;. The connector link B;H; is either in compression or tension, as illustrated in Figure 2.24. This is essentially a forward static analysis of a single connec- tor which identifies the states of compression and tension in a connector. Conversely, when an external force f; is applied at point C;, then an equalStatics of a parallel manipulator 71 Busy © rf (a) Compression (b) Tension Figure 2.24 A static analysis of a single connector. and opposite force f; must be generated in the cylinder which acts at point E;. The connector link C;E; can be either in compression or tension, as illustrated. Finally, when the external force f; is applied at point C;, an equal and opposite reaction force f; is transmitted from the fixed platform at point B; to the con- nector link B,H;, which can be in a state of compression or tension, as shown in Figure 2.24, This is essentially a reverse static analysis of a single connec- tor which identifies the states of compression and tension in the connector. EXERCISE) 29 creas 1 (a) Determine the unitized coordinates of the lines that join the con- nectors BiCi, ByC>, and B3C3 using the point By (Bz) as your reference (see Exr. Fig. 2.3(1)). Hence, determine (2.67), = jA for the truss. (b) A vertical force of 10 Ibf is applied through point C). Compute the equilibrating connector forces fi, f2, and f3, and state whether each connector is in compression or tension. (© A vertical force of 10 Ibf is applied through point C2, and the force through point C; is removed. Compute the equilibrating connector forces and state whether each connector is in compression or tension. Confirm this result by choosing your reference point at C2 (C3). 2. (a) Determine the unitized coordinates for the lines that join the con- nectors ByC}, ByC>, and B3C3 using the point B, (Bz) as your reference (see Exr. Fig. 2.3(2)). Hence, determine (2.67), w = jA for the truss. (b) A vertical force of 5 lbf is applied one inch to the right of point C.Planar statics 72 B,, Bz Exercise Figure 2.3(2) Compute the equilibrating connector forces and state whether each connector is in compression or tension. (©) Compute the equilibrating connector forces when the vertical force of 5 Ibf is applied at C; and then at C2(C3). Choose your reference points at B,(B2), then C2(C3). 2.8 The geometrical meaning of j~? The inverse of cy C2 63 F=] 1 52 53 Pi Pz P3.Geometrical meaning of j-* 73 can be determined by constructing the matrix, (saps — S3p2)— (S3p1 — Sips) (SP2 — S2p1) J =| —(2p3 — esp2) —(capx — c1p3) —(cup2— exp) »_— (2.71) (€253 — €352) —— (€381 — 183) (C182 — C251) where the elements of j’ are the signed minors of j. The inverse j~! can thus be expressed in the form = j’Tidet j, (2.72) where C1 C2 3 detj=|s: s2 33], Pi P2 P3 and j’7 denotes the transpose of j’. The geometrical meaning of j~! can be understood by solving sequentially the equations for the connector lines, which can expressed in the form cy — sx + py = 0, (2.73) coy — ox + pr = 0, (2.74) cay — 53x + p3 = 0. (2.75) The solution of (2.73) and (2.74) for the coordinates (x12, y12) of the point of intersection of the lines $, and $2, using Grassmann’s expansion (2.19), yields (see Fig. 2.25) M12? X12 1 = (S1p2 — Sopa): (C12 — €2pr) : (C182 — C281). (2.76) Analogously, Ya: X31: 1 = (931 — $13) : — (capi — caps): (381 — €183), (2.77) and ‘Yo3 1 —X93 21 = (823 — 53p2); —(C2p3 — Cap2) : (C283 — €382). (2.78) Chapter 3 will show that the array {y, —x, 1} consists of the coordinates for a line $ through a point with coordinates (x, y) and which is perpendic- ular to the XY plane. Hence, the columns in (2.71) are, to a scalar multiple,Planar statics 74 Figure 2.25 Skeleton form of a parallel manipulator illustrating the lines associated with the connectors. the unitized homogeneous line coordinates of the lines $23, $31, and $12 through the points labeled 23, 13, and 12, which are perpendicular to the plane of the page (see Fig. 2.25). Therefore, (sep3 — s3pa) 23 i‘ ~ (Copa — €3p2) | = $3-2| —X23 | = 53-2 Sos, (2.79) (e283 — €382) 1 (sap1 — sips) 31 - — (capi — €1ps) | = 81-3 | x31 | = 51-3 Sar, (2.80) (c381 — €183). 1 (sip2 — Sop) v2 : (capa — cypy) | = 82-1 | x12 | = 82-1 Sir, (2.81) (c1s2 — C281) 1 where the trigonometrical identity cjs; — cjs; = s;-; has been used. Further- more, j~! can be expressed in the abbreviated form 53-2 Sf = | sy 81, [det 7. (2.82) 52-1 StSingularity configurations of a parallel manipulator 75 The rows of j~! are therefore, to a scalar multiple, the unitized coordinates for the lines $23, $31, and $12. The expansion of det j from the first column yields det j = c1(s2p3 — s3p2) — $1(Cop3 — ¢3p2) + Prle2s3 — €382), (2.83) which, from (2.79), can be expressed in the form det j = 87 53-2 S43 = 83-2 ST,81. (2.84) Analogously, expansion from the second and third columns yields det j = SF y-3 ST = 1-3 E50, (2.85) and det j = 87 52-1 Sy2 = 5-1 S783, (2.86) The substitution of (2.82) into the right side of (2.70) yields fi 3-2 So r A=] fe] =| 91-3 Sar] / det j. (2.87) A 2-1 S12 Therefore, from (2.87) and (2.84)-(2.86): fi = 84, w/ ST, (2.88) fr = SH! She, (2.89) fi = STI STS, (2.90) The geometrical meaning of these expressions will be made clear in Section 4.4 2.9 Singularity configurations of a parallel manipulator It is clear that when the rank of j is less than 3, then det j = 0 (see equation 2.68) and hence, j~! does not exist. When this occurs, the manipu- lator is in a special or singularity configuration. It is well known that when cy C2 C3 detj=|5 s2 53) =0, Pil P2 P3Planar statics 76 oO Figure 2.26 Singularity configuration; the connector forces are con- current. then the three lines are concurrent, i.e., they all meet in a finite or an infinite point. A configuration for Q finite is illustrated in Figure 2.26. It is easy to deduce the det j = 0, for if the origin O of the XY coordinate system were transformed to Q, then py = p2 = p3 = 0. Also, the lines of ac- tion of the forces pass through Q, and the three forces are thus linearly de- pendent (see Section 2.4). Consider that the origin for a coordinate system is located at Q and a force with coordinates w= (L, M; R) is applied to the top platform. It follows that L cr cz es] [fi Ml=|5 % 8 ||fAl. (291) R oo Odlfs It is not possible to solve (2.91) for fi, 2, and f;; clearly there is a couple, with magnitude R acting, which cannot be equilibrated by the connector forces generated. We will show in Section 4.6 that the top platform has an uncon- trollable instant mobility, that is, an instantaneous rotation about Q. A singular configuration occurs when all three connector forces are on parallel lines (or the point of intersection Q is at infinity). The geometry isSingularity configurations of a parallel manipulator 77 oT 3 Figure 2.27 Singularity configuration; the connector forces are par- allel. illustrated in Figure 2.27. The slopes of the lines of action of the three forces are the same and equal to tan A. Therefore, SD - 2 = BH any, (2.92) et ee e2 C3. The first two rows of det j are identical to a scalar multiple of tan A, and thus det j = 0. The three forces belong to a parallel pencil of lines and are lin- early dependent (see Section 2.4). Consider that a coordinate system is cho- sen at a point O with the X axis drawn parallel to the legs, and a force w = {L, M; R} is applied to the top platform. It follows that L 11 177A M|=/]0 0 OJ fF]. (2.93) R 0 p pIlh It is not possible to solve (2.93) for fi, fs, and fs; clearly, any force parallel to the ¥ axis cannot be equilibrated by the connector forces generated. Sec- tion 4.6 shows that the top platform has an uncontrollable instant mobility, that is, an instantaneous translation parallel to the Y axis. A singular position occurs when any two connector forces lie on the same line. Figure 2.28 illustrates the case for f; = Af|, where A # 0. (Clearly, two more cases occur when f; = Af; and f, = Af3, which are analogous to the geometry illustrated by the figure.) The first two columns of det j are iden-Planar statics 78 Figure 2.28 Singularity configuration; a pair of connector forces whose lines of action coalesce. 7 i et 3 Figure 2.29 In-parallel manipulator with special geometry. tical to a scalar multiple and hence det j = 0. It is clear from Figure 2.28 that there are two independent connector forces fj + f) = (1 + A)fi and f3. The lines of action of these forces meet at a point Q, and their resultant thus lies on the pencil of lines through Q. Any force applied to the top platform can only be equilibrated if its line of action passes through Q. Finally, consider a parallel manipulator specialized to the extent that all three points of connection of both the moving and fixed platforms are collinear (see Fig. 2.29). Three types of singular configurations occur, which are illustrated by Figure 2.30. It is interesting to note that for this special geometry all three connector forces can become collinear, simul- taneously.Singularity configurations of a parallel manipulator 79 I rt ce T 2 a Figure 2.30 Three singularity configurations. ‘We may conclude that singular or special configurations of in-parallel pla- nar manipulators can be described elegantly by the geometry of pencils of concurrent or parallel lines. Other special cases occur when two or more lines coalesce. —— EXERCISE 2.4 Exercise Figure 2.4 illustrates an in-parallel manipulator with special geometry (see also Fig. 2.29) with the moving platform in an initial position for which the coordinates of points 1, 2, and 3 are, respectively, (0, 3), (0.4V2, 2.4), and (0.95 V2, 1.7). 1. Write the unitized coordinates [c, s, p]" of the force f which passes through point 2 and is parallel to the x axis as shown in the figure. 2. Give the platform self-parallel displacements in increments of 0.1 inches away from the initial position up to 2.9 inches. For each increment compute the coor- dinates of the points 1, 2, and 3, i.e., (x1, 3), (X2n, 2.4), and (x3, 1.7) for n = 0, 1,...,.29. 3. Compute the corresponding connector lengths (Cin, €2n, €3n)- 4. Use Grassmann’s method to compute the corresponding sets of unitized line connector coordinates [cj,, Sins Pin] for i = 1, 2, 3, and n = 0, 1,..., 29. Hence, compute the matrix [j] for each position and, correspondingly, compute the in- verse [j]“}.Planar statics 80 Exercise Figure 2.4 5. Compute the resulting leg forces for each position using Lt You may assume that f= 1 unit of force. 6. Plot the equilibrating force in each of the three connectors against the hori- zontal position x. Label each curve and state whether a connector is in tension or compression. 2.10 Statically redundant parallel manipulators Let us consider the static analysis of a platform with four connec- tors (see Fig. 2.31). Here the force on the top platform and the connector forces are related by C1 C2 C3 C4 w=fi| | +h] x2) +A] 53 [+A] sa |. (2.94) Pi P2 P3 Pa. Equation (2.94) can be used to compute the resultant of the connector forces. However, the reverse static analysis is not possible. It is not possible to com-Statically redundant parallel manipulators 81 r 4 Figure 2.31 A redundant parallel manipulator. pute a unique set of magnitudes f;, fo, f5, and f, which are superabundant by one. Equation 2.94 can be expressed in the matrix form by waja, (2.95) where CG. C2 C3 Og S182 53 Sd, (2.96) Pi P2 P3 Pa and A= - (2.97) Clearly, j is noninvertible. Systems with four or more connectors are stati- cally indeterminate, and their solution is beyond the scope of this text.3 First-order instantaneous planar kinematics 3.1 The definition of a rotor ‘We will consider a rigid lamina that moves on a fixed reference plane and instantaneously rotates about a given fixed axis parallel to a unit vector k with an angular speed w. The angular velocity is given by w = wk (see Fig. 3.1). All points on the lamina which do not lie on the axis of rotation are moving on circular paths, and the instantaneous velocity of a point spec- ified by a vector r drawn from a point Q in the fixed plane which lies on the axis of rotation is given by v=@Xr. GB.) In what follows, we choose some reference point O in the fixed plane which is not on the axis of rotation (see Fig. 3.2), and determine the veloc- ity v, of a point in the moving lamina instantaneously coincident with O. The substitutions of v = v, and r = —r, in the left and right sides of (3.1) yield Vo = @ X (=I) = fy X @, (3.2) where v, is the required velocity of the point on the moving lamina instan- taneously coincident with the point O in the frame of reference. We now define an instantaneous rotation as a rotor, which has the coor- dinates 7 = {a vo}, (3.3) 82Coordinates of a line parallel to the Z axis 83 o=ok rotation axis Figure 3.1 Instantaneous rotation. o=ok Figure 3.2 Rotation with the reference point at O. where vy = rr, X @. This is analogous to the representation of the coordinates of a force w = {fi eo}, (3.4) where ¢, is a couple and ¢, = r, X f. A comparison of (3.3) and (3.4) shows that the vector forms of a force and a rotor are analogous. The direct analogy between the statics and instanta- neous kinematics of a rigid body is becoming apparent. In statics a directed line segment represents the “rectilinear” concept of force, whereas in kine- matics a directed line segment represents the “circular” concept of a rotor. 3.2 The coordinates of a line parallel to the Z axis Any directed line segment parallel to the Z axis can be expressed by S = Nk, (3.5) where |S| = N. Clearly, Nk is a free vector and it can be associated with anyFirst-order instantaneous planar kinematics 84 Zz x Figure 3.3 A line parallel to the Z axis. line drawn parallel or antiparallel (NW < 0) to the Z axis. Assume that the line segment passes through the point U with x and y coordinates (xo, Yo). The moment of the line segment about the origin is given by r, X S = (xi + Yoj) X NK = QoN)i + (—x0N)j, 3.6) which can be expressed in the form r, X S = Pit Qi, 3.7) where P = y,N and Q = —x,N are the components of the moment about the X and Y axes, respectively. The three numbers N, P, and Q are used as the coordinates of the line. They are homogeneous: MN, AP, and AQ (where A is a nonzero scalar) de- termine the same line. The units of N, P, and Q are not consistent. N has di- mension (length)!, while P and Q have dimensions of area, namely (length). Because of this inconsistency in dimensions, the coordinates are represented by the ordered triple of real numbers {N; P, Q} with the semicolon separat- ing N from P and Q. Furthermore, when N = 1, P= yo, and Q@ = —xo. In other words, the coordinates for a directed line segment of unit length par- allel to the Z axis and passing through a point with coordinates (Xo, Yo) are {1; Yo. —Xo}. Figure 3.3 shows that the moments of the unit vector S about the X and ¥ axes are y, and —x, respectively. If the length scales on the axesVector representation of a rotor 85 Figure 3.4 A line in three dimensions. were altered from say, inches to millimeters, then to get the same line, P and Q would need to be larger by a factor of 25.4. The reader may recall that the homogeneous coordinates for a line in the XY plane were expressed by {L, M; R} (see Section 2.2). The coordinates (L, M) de- termined the direction of the line, and R, the moment of a line segment about the Z axis. It has been demonstrated that for a line parallel to the Z axis the direction ratio is N, and the coordinates (P, Q) are moments about the X and Y axes, re- spectively. These two representations are consistent and they can be expressed as {L, M, 0; 0, 0, R} Line in XY plane, {0, 0, N; P, Q,0} Line parallel to Z axis (perpendicular to the XY plane). In fact, the homogeneous coordinates for a line in three-dimensional space are the sextuple {L, M, N; P, Q, R} (see Fig. 3.4). 3.3 A.useful vector representation of a rotor Figure 3.5 illustrates a lamina rotating about an axis through point 1 and normal to the page, the XY plane, with an instantaneous angularFirst-order instantaneous planar kinematics 86 Figure 3.5 A rotor modeled by a revolute pair. speed w. The instantaneous motion can be represented by a rotor with co- ordinates ? = (ew; vo}, (eq. 3.3) where v, = 1 X is the velocity of a point in the lamina coincident with a reference point O fixed in the reference XY plane. This representation can be related directly to the line $ passing through point 1 which has unit coordinates § = {k; S,}. (3.8) Clearly, |k| = 1, and S, = 1, X kK = yoi — xoj. In addition, @ = ok, Vo = (wyo)i + (— axo)j = val + vyJ, 3.9) where v, = wy, and v, = — wx, are the components of v, in the X and ¥ di- rections, From (3.3) and (3.9): 2 = @ {k; Yol — xoj}- (3.10) The coordinates for the rotor are thus {@; wy,, —@x,} or {@; vz, vy}. The substitution of (3.8) into the right side of (3.10) yields f=o8. (3.11)Infinitesimal displacements of a rigid lamina 87 The rotor is thus expressed as a scalar multiple of the unit line vector $, de- fined by (3.8). 3.4 — Infinitesimal displacements of a rigid lamina Assume now that the rigid lamina (see Fig. 3.1) moves on the fixed reference plane by undergoing a small rotation 5¢ about an axis parallel to the unit vector k (see Fig. 3.6). The tangential displacement of a point on the moving lamina coincident with a point O in the frame of reference (see also Fig. 3.6) is given approximately by Oo = 1, X Sok. (3.12) The tangential and rotational displacements 5r, and 5¢ are easy to relate to the tangential and angular velocities v, and w by dividing both sides of (3.12) by a small time increment 6t. As dt O, Oe _, do _ 8), WK = ok = Shim vo and a a * ok = o, and (3.12) can be written in the form v, = r, X @, which is precisely (3.2). Now To = Xoi + yoi, (3.13) and therefore fo = (Kol + Vos) X 8Gk = Shyoi — Sbxoj, (3.14) which can also be expressed in the form Oy = Sxoi + Syoj, (3.15) where 8x, = Shy, Yo = —Sbxo (3.16) are the components of Sr, in the X and Y directions. The infinitesimal displacement will be denoted by 8d = {(K, yoi — rol}, G17) which is completely analogous to the twist # = wfk; yoi — Xai). (eq. 3.10) The coordinates for dd are thus {86; 6pyo, — 8x0} or (8d; Stuy Byo)-First-order instantaneous planar kinematics 88 Figure 3.6 Small rotation 5¢. Figure 3.7 A pure translation. 3.5 A representation of pure translation Assume that the rigid lamina (Fig. 3.7) is moving with a pure trans- lational velocity v,, that is, every point in the lamina moves with this veloc- ity. Clearly, @ = O and the representation of this motion is given by t= (0; vo}, (3.18) where {O; v,} is a free vector and not a line-bound vector. The instantaneous translation v, of the lamina, however, can be considered a rotor with infini- tesimal magnitude 5 whose axis k is perpendicular to the plane of transla-Ray and axis coordinates of a line and a rotor 89 tion. If we write r, = ‘ol, Where u is a unit vector parallel to r,, then v, = (Swr,)(u X k) is finite as 8@— 0 and 1, > ©. This is analogous to the rep- resentation of a pure couple (see Section 2.3). 3.6 Ray and axis coordinates of a line and a rotor Thus far, the coordinates § for a line $ have been expressed by the ordered pair of vectors {S; S,} and 8 = {S;S,]. (3.19) In this representation, the coordinates § are defined as ray coordinates be- cause the line that can be drawn through the two points labeled 1 and 2 may be imagined as an actual ray or narrow beam of light originating at a point (see Fig. 3.8). The coordinates for the same line $ can be expressed equally by the or- dered pair of vectors {S,; S} and denoted by 5, where $= {S,; S}. (3.20) In this representation, the coordinates § are defined as axis coordinates be- cause the line $ is considered as the meet of two planes (see Fig. 3.8). A pen- cil of planes could be drawn through $, which can be considered the axis of the pencil. (The lower case § and the upper case § distinguish the two coor- dinate representations of the same line $.) It is popular to express the coordinates for a rotor ¢ in axis coordinates. The ray coordinates for r are T= {@; Vo}, (eq. 3.3) or f=05, (eq. 3.11) where $= (8; S,}, (eq. 3.19) and, similarly, the axis coordinates T for the same rotor t are T= {v,; w), (3.21)First-order instantaneous planar kinematics 90 Figure 3.8 A line $ defined by two points or two planes. or ~> i & (3.22) where $= (S,; S}. (eq. 3.20) From (3.10) the axis coordinates for a rotor perpendicular to the XY plane are T= o {yoi — xoj; k}. (3.23) The axis coordinates for the rotor are thus {wyo, —@xo; w} oF {Vx, Vy; w}, and the axis coordinates for a pure translation are {v,, vy; 0}. Analogously, the axis coordinates 8D for an infinitesimal rotation can be written in the form SD = 56 {yoi — Kol k}. (3.24) Divide the left and right sides of (3.24) by df and write 2. 8D 7 Poy mt om to yield (3.23). The axis coordinates are thus expressible in the forms { d¢y,, —68¢x.; 5p} or { 5x4, Syo; 5h} and the coordinates for an infinitesimal trans- lation are { 8xo, d5yo; 0}. From here on, the coordinates for lines and rotors perpendicular to the XY plane will be expressed in axis coordinates.Translation and rotation of coordinate systems 91 3.7 ‘Translation and rotation of coordinate systems The Pliicker coordinates of a unit line bound vector perpendicular to the XY plane though point U with coordinates (x, y) are {0, y, —x; 0, 0, 1}, which will be written in the abbreviated form {y, x3 1}. (3.25) The coordinates after the translation are {y’, —x'5 1}. (3.26) The translation is illustrated by Figure 3.9 and x=x' +a, y= tb. (3.27) It is clear that point U does not change under this transformation. There- fore, neither does the directed line segment pointing out of the page through the point U nor the line of unlimited length passing through U. All these geo- metric elements, point, directed line segment, and line, are therefore invari- ant with a translation of coordinate system, although the actual coordinates in the translated system are different. The arrangement of (3.27) in the form of (3.25) yields y=y'+b, -x=-x'-a, 1=1, (3.28) which can be expressed in matrix form as y 10 6 -x}=|0 1 -a (3.29) 1 001 The coordinate {0, 0; 1} remains unchanged under the translation and it is therefore independent of the coordinate system. It defines a free vector k parallel to the Z axis and its value remains unchanged under parallel dis- placement. All three coordinates {y, —x; 1} define a line bound vector and the coordinates {y’, —x’; 1} define precisely the identical line bound vector, the line segment (or rotor) in the second coordinate system. As stated before, the coordinates {y, —x; 0} for the moment change are origin dependent. How- ever, since y’ and —x' are solely functions of b and a, they quantify the mo- ment about a different origin O'.First-order instantaneous planar kinematics 92 Oo ~X Figure 3.9 A translation of a coordinate system. The rotation of a rectangular coordinate system through an angle ¢ is il- lustrated in Figure 3.10. From (2.42), x’ = x'cos @— y" sin d, (3.30) y' = x'sin 6 + y" cos $. Therefore y' = y" cos @ — (—x’) sin g, —x' = y’ sin @ + (—x’) cos ¢, 1=1, (3.31) which can be expressed in the matrix form y cos@ ~—sing 0 y" —x'|=|sind cosd Of] -x"]. (3.32) 1 0 0 1 1 The matrix equation (3.32) expresses the coordinates of the rotor in the old coordinate system in terms of the coordinates of the rotor in the new coor- dinate system. For this transformation the coordinate {0, 0; 1} remains un- changed. The coordinates {y’, —x’; 0} in the new coordinate system depend solely upon the original coordinates {y', —x'; 0}, and the relationship of y”, —x" to y’, —x’ is the same as (x”, y”) to (x', y’).Relationship betwen [e] and [E] 93 Figure 3.10 A rotation of a coordinate system. 3.8 — Induced line (rotor) transformation of the Euclidean group For both translation and rotation of the coordinate system, the in- duced line transformation obtained from (3.29) and (3.32) can be expressed in the form y y —x | = [El] -x" |, (3.33) 1 1 where 10 6b Cote SiO (tieteteet desseet 2 [E}=|]0 1 -a]|s ¢ O}=]s ec —a}. (3.34) 00 1 0 0 1 00 1 3.9 Relationship between [e] and [E] In Chapter 2 we derived and expressed the transformation [e] for a force or line in the XY plane under the Euclidean group in the form c -s 0 [el = s c oO}. (eq. 2.53) as—be act+bs 1First-order instantaneous planar kinematics 94 Forming the inverse of [e] yields Cars eOe las tlt=|s c -al. (3.35) 0 0 1 Clearly, from (3.34) the right side of (3.35) is [E]’. Hence lel = (ey. aa Analogously, forming [E]~! using (3.34) and comparing with (2.53) yields el” = [E". (3.37) EXERCISE 3.1 1. Determine the coordinates of the points of intersection of the following lines with the Z = 0 plane: {12, 16; 4}, {12, —16; 4}, {—12, 16; 4}, {12, 16; —4}, and {—12, —16; —4}. 2. Determine the 3 X 3 matrix representations [A] and [B] of the induced line transformation for a translation and rotation of an XY coordinate system. Obtain [A] and [B] for a = 2 inches, b = 3 inches, and @ = 60 degrees. Finally, deter- mine the product [A] [B]. 3.10 The first-order instantaneous kinematics of a 3R serial manipulator The end link a4 that carries the end effector is connected to the ground via three serially connected revolute joints. Assume at the outset that the geometric configuration (4), 62, 63) is known (see Fig. 3.11). Assume that three rotors with magnitudes @), 2, and w3 are generated in each joint about the lines $1, $2, and $3, which are parallel to the Z axis. It is important to recognize that @, is the angular velocity of link a)2 with re- spect to the ground, w» is the angular velocity of link a3 relative to link aj, and «3 is the angular velocity of link a34 with respect to link a)3. The end effector will move on some rotor T whose coordinates are given by T=Ti+h+h, (3.38) which can be expressed in the alternative formFirst-order kinematics of a 3R serial manipulator 95 Figure 3.11 A 3R serial manipulator. [=J-[e]+ [+o -[: ]= od + oS + 08 ‘0 o Therefore, from (3.40): T T fia) ‘9 @ = wk + ok + w3k, (3.41) and Vo = @S8o1 + w28o2 + «38,3. (3.42) From (3.41): o =, + a t+ 03. (3.43) From (3.42): $= 181+ 252+ 2,5. (3.44) o @ @ In the previous equations T,, ty, and Ty represent, respectively, the coordi- nates of the motions of link a1 with respect to the ground, link a3 with respectFirst-order instantaneous planar kinematics 96 to link aj, and link a34 with respect to a3. The resultant T thus represents the coordinates of link a34 with respect to the ground. The translation velocities v1, Vo2, and V3, are, respectively, the velocities of points in the laminas containing links a2, a3, and a34 which are coincident with the reference point O. It is often convenient to express (3.40) in the expanded form v1 ye %3 T= @ | —x | + @2 | —x2 | + @3 | —x3 |, (3.45) i 1 1 where » (= 1, 2,3) 1 are the Pliicker coordinates of the lines that pass through the joint axes. Equation 3.45 can be expressed in matrix form by T= Jy, (3.46) where vy y2 3 J=|-m -m2 -x3 (3.47) 1 1 1 is a3 X 3 matrix, and Vox a T=| vy y=| or (3.48) o 3, are 3 X 1 column vectors. When the magnitudes «), w, and w3 are specified it is a simple matter to compute the coordinates T of the resultant rotor using (3.40). This is called the forward kinematic analysis. Conversely, when the end effector moves on some rotor with coordinates 7, we need to determine the magnitudes @, w», and @3. The solution is easy to obtain from (3.46) and yas F, (3.49)Geometrical meaning of J-* 97 where J~! is the inverse of J. This is called the reverse velocity analysis. Finally, substituting ~_) 8D — im 86 T=jn sy and «= jim into (3.46) and canceling throughout by 6 yields 8D = J60, (3.50) where Xo 801 8D = | by. |, 50 = | 56, |, (3.51) 86 505 which are analogous to (3.46) and (3.48). Equation 3.50 thus relates the in- finitesimal joint displacements to the infinitesimal motion of the end effec- tor. Therefore, 60=J-'8, (3.52) which is analogous to (3.49). 3.11 The geometrical meaning of J-1 Now from (3.47): x3 — X2 x1 — X3 mx |F JT=) y3-y2 Ya 3 Y2— Yi | det J X23 — X3Y2 X3Vi — XiY3 X12 — ¥2V1. 423 5 =| ay $3, |/det J, (3.53) 412 Sh where by Grassmann (see Section 2.1) $23, §31, and §,2 are the column vec- tors formed by the unitized coordinates for the lines joining the points 2-3, 3-1, and 1-2 (see Fig. 3.12). For instance, employing the usual notation Sh, = [coy 825 pa] and [(x3 — x2), (3 ~ ¥2)s Grays — x3Y2)] = a3 S23, where ay3 = {(x3 — 2)? + (93 — y2)?}"? is the distance between joints 2 and 3. Further-First-order instantaneous planar kinematics 98 Figure 3.12 The lines associated with a 3R serial manipulator. more, det J can be expressed in three ways. Expansion from the first column of the right side of (3.47) yields det J = yi(%3 — x2) — x13 — y2) + 1 (aay3 — xay2), (3.54) which can be expressed in the form det J = 8Tay3803 = ans87, $1. (3.55) Analogously, expansion from the second and third columns of the right side of (3.47) yields det J = Sfay8s1 = azi8, $2, (3.56) and det J = S$ay812 = ay287, $3. (3.57) The substitution of (3.53) into (3.49) yields @ ans5y . y=] or] =|] asf, [/aets |? (3.58) Os, 41281) Therefore, from (3.58) and (3.55)-(3.57): ow = 3, F157,81, (3.59)Geometrical meaning of J-7 99 ow = 54, T1578, (3.60) 3 = 57, F157,8;, (3.61) Chapter 4 shows that (3.59)-(3.61) can be obtained directly from (3.40) by forming so-called reciprocal products. EXERCISE 3.2 1. (a) A lamina is instantaneously rotating about a point C with coordi- nates (x,, y<). The velocity v, of a point P is known. The coordinates of P are (2p, Yp). Obtain an expression for the angular velocity w in terms of vp and the coordinates of C and P. Hence, derive an expression for the velocity v, of a point in the Jamina coincident with the reference point O and draw its direction on Ex- ercise Figure 3.2(1a). (b) A lamina is instantaneously rotating about a point C with coordi- nates (x,, y.), Where x. = 0 and y. = 2 inches. The velocity v, of a point P is vp = | inJsec. (see Exr. Fig. 3.2(1b)). The coordinates of P are (xp, yp), Where Xp = 1 inch and yp = 2 inch. Use the expression from part (a) to determine w and the velocity v, of a point in the lamina coincident with the reference point O. Compute the joint velocities @, @, and @; for the following 3R manipula- tor. (Note: For a single calculation it is as easy to solve three simultaneous equa- tions for @, @, and @s as it is to invert a 3 X 3 matrix.) 2. Compute the joint velocities @;, v2, and 3 of the RPR manipulator for each end-effector velocity (see Exr. Fig. 3.2(2)): . fol]. yo 0 T= 00), T=u]0], T=yfr]. 1 0. 0 Write a short comment on each result; explain how it makes sense physically. 3. (a) Write the coordinates 7 for the instant motion of the end effector of the 3R robot analyzed in 1(b). (b) Use Grassmann (see equation (2.9)) to determine the coordinates of S23, $31, and $12 of the lines $23, $3), and $12 for the 3R robot shown in Ex- ercise Figure 3.2(3). (c) Write the coordinates $1, 53, and $3 of the lines $1, $2, and $3 through the points (0, 0), (3, 0), and (2, 2) which are perpendicular to the XY plane.First-order instantaneous planar kinematics 100 ‘Pe oO Exercise Figure 3.2(1a) 1 inksec Exercise Figure 3.2(1b) (a) Show that 5%, 2 = S23 $5 = 0, 8%, $3 = 37, $1 = 0, and $7, $1 = 51, Sy = 0. (e) Compute @, @2, and «3 using (3.59)-(3.61). 4. It has been shown that for a 3R manipulator: det J = a1257,5, (see equation (3.57)) which is twice the area of the triangle 123 (see also Exr. Fig. 3.2(4)). (a) Deduce that det J = a)2a93 sin 6). (b) Deduce that the maximum value for det J is det Jinax = 412223 and that A = det J/det Jmax = sin 0. (c) Sketch the graph A = sin 6, for 0 = @ = 360 degrees. Draw the pairs configurations of the manipulator for A = 0 and A = +1. Note: A is independent of the manipulator dimensions.Geometrical meaning of J~* 101 25,0) Exercise Figure 3.2(2) Exercise Figure 3.2(4)First-order instantaneous planar kinematics 102 Figure 3.13 Singularity configurations of a 3R serial manipulator. 3.12 Singularity configurations of a serial manipulator Choose the origin of a coordinate system to lie on the first joint axis, x1 = y1 = 0. From (3.47) det J = 0 when 2 = Bwana, (3.62) ™ % as illustrated by Figure 3.13. Points 2 and 3 lie on a line through the origin which is defined as an extreme distance line. When a reference point Q in the end effector lies on this line it is at an extreme distance from the first grounded joint. Rotating the coordinate system so that the rotor axes lie on the X axis yields (see also Fig. 3.14) y 0 0 (Oo |fa T=w|-x|=]0 -x. -x3|] «|. (3.63) 1 1 1 1 JLos It is not possible for the end effector to move on a rotor T with coordinatesSingularity configurations of a serial manipulator 103 Figure 3.14 Point U must be on X axis. w{y, ~x; 1} whose axis passes through the point U, no matter what the in- tensities «1, @2, and w3 of the three rotors T;, T>, and T; are. Clearly, from (3.63) wy = 0, and since y # 0, then w = 0. ‘The singularity configurations of 2R-P and R-2P manipulators are dis- cussed in question 2 of Exercise 3.3. EXERCISE 3.3 1. (a) Use parallel projection to obtain expressions for the coordinates (x2, y2), (3, y3) Of the points 2 and 3 of the 2R planar manipulator (see Exr. Fig. 3.3(1)). (b) Perform a complete reverse displacement analysis, ie., obtain ex- pressions for cos 6. See also subsection 1.4.1 (cos 6;, sin 1). 2. The end effector of the 3R manipulator is moving with a pure translational constant velocity v parallel to the x axis, as illustrated in Exercise Figure 3.3(2). Give the end-effector self-parallel displacements in increments 6 = 0.25 (0 = 5S 2.25) and 5 = 0.025 (2.25 < 6 = 2.5) of 0.2 feet away from the initial po- sition (x3, y3) = (0, 2.5) (ain = 2 ft, ao3 = 1.5 ft). (a) For each set of coordinates (x3, y3) compute corresponding sets of and 6, using the reverse analysis in question 1. Then compute the corre- sponding values for the coordinates (x2, y2). (b) Compute the matrix for each increment and correspondingly compute J~!.First-order instantaneous planar kinematics 104 Exercise Figure 3.3(2) (c) Compute the corresponding angular velocities of the joints ft A wo |= JF, wy. using a value of vy = | in./sec. (d) Plot @, a, and 3 against x. 3. (a) Write the matrix J for the PRR manipulator (see Exr. Fig. 3.3(3a)).Singularity configurations of a serial manipulator 105 7, NN, (X35 3) a x o (%2, 0) Exercise Figure 3.3(3a) Exercise Figure 3.3(3b) Show that det J = 0 when x) = x3. Draw this configuration and comment on your result. (b) Write the matrix J for the RPR manipulator shown in Exercise Fig- ure 3.3(3b). Verify that x3 = a2 Cos 0 + Sp Cos ¥, ‘y3 = ap sin 0, + S sin y, y= + W2. Use these expressions to show that det J = 0 when S> = 0. Draw this configu- ration and comment on your results. (©) Write the matrix J for the RRP manipulator shown in Exercise Fig- ure 3,3(3c). Verify the expressions x2 = ay2 cos 64, y2 = a12 sin , y= 01 +First-order instantaneous planar kinematics 106 (25 Y2) Exercise Figure 3.3(3c) 6, + 7/2, and show that det J = 0 when sin @ = 0. Draw these two configura- tions and comment on your results. (d) Write the J matrices for the 2P-R manipulators and obtain the sin- gularity condition for the PRP manipulator. 3.13 Kinematics of redundant serial manipulators A serial manipulator with four joints, that is, a PRRR manipulator, is shown in Figure 3.15. Assume that the end effector moves on a rotor whose axis passes through a point with coordinates (x, y); then the instantaneous kinematics of the manipulator requires the solution of ‘i y 1 0 M3 M4 T= @| -x| =v] 0] + @ | —x2 | + @3 | —x3 | +o] —xq]. (3.64) 1 0. He 1 1 Clearly, (3.64) can be used to compute the resultant rotor for a given set of joint speeds. However the inverse velocity problem does not have a unique solution: It is not possible, in general, to solve uniquely for {v1, @2, 3, «4} because the matrix 1 0 ys YM J=)0 -x. -x3 —x4 (3.65) Oo 4. 1 1 is noninvertible. The inverse velocity analysis of serial manipulators with four or more rotors is indeterminate.First-order kinematics of a closed-loop 4R mechanism 107 Figure 3.15 A PRRR redundant manipulator. 3.14 First-order instantaneous kinematics of a closed-loop 4A mechanism The twist equation for the closed-loop 4R mechanism (see Fig. 3.16) can be expressed as Ty +t,+0,+ 7, =0, (3.66) where a T ip and ts are the Coordinates of the rotors through points 4, 1, 2, and 3 which have the coordinates (0, 0), (x1, 0), (x2, y2), and (x3, y3), re- spectively. Therefore 0 0 ye Bz 0 ws} 0} +1] —m | +o] —2 | +3] -23 |=] 0]. 6.67) 1 1 1 1 0 Note that the coordinates ty, ty, Ts and iT; represent, respectively, the rela- tive motions of link a1) with respect to the frame a4), a3 with respect to a2, 434 with respect to a3, and the relative motion of the frame a4, with respect 10 434. In other words, @), @, and w3 are, respectively, the angular veloci- ties of link a2 with respect to the frame a41, a3 with respect to a12, and a34 with respect to a23, whereas wy is the angular velocity of the frame a4, mea- sured relative to a34. Usually, the input angular velocity o, of the movingFirst-order instantaneous planar kinematics 108 Figure 3.16 A planar 4R mechanism. link a34 with respect to the frame a4) is specified. Clearly, in (3.67) and in the following matrix form of (3.67) w4 = — w,. 0 0 yo Ilo oy) O]= —|—x x2 x3] ] oe |. (3.68) 1 1 ot 1 JLes. This equation can be used to compute the rotor speeds @, @, and w; for any specified input rotor speed w, of link aaq with respect to a4), with w, = —w, provided that OF v2 3 det J=]-x, x —2x5] +0. (3.69) eee The expansion of (3.69) from the first row shows that det J = 0 when —2— = —3_ (= tan a), (3.70) 7X2 My HB for which the input a34 is in a stationary position relative to the reference frame, and w, = 0, as illustrated by Figure 3.17. It is interesting to note that although the 4R mechanism is in a limit con- figuration and it has degenerated into a triangle, namely a structure, the axes of the three joints 1, 2, and 3 are in the same plane perpendicular to the XYFirst-order kinematics of a closed-loop 4R mechanism 109 Figure 3.17 Stationary configuration of the input crank. T; T, G23 2 ap a o® © SE Figure 3.18 An apparent structure which has infinitesimal mobility. plane, and hence, they are linearly dependent. For convenience, a new XY coordinate system is chosen, as illustrated in Figure 3.18. Since w4 = 0, then in this new coordinate system ofhe[eket Ee) om Equation 3,71 is obtained by substituting y2 = x3 = y3 = 0 in (3.67). All three rotors span a two-space and belong to a two-system. For example, if one specifies the angular velocity w3, then the angular velocities w, and w can be computed from (3.71). In other words, this apparent structure has instant mobility, This infinitesimal mobility can be detected by pushing up or down on the assemblage. Figure 3.19 illustrates a 4R mechanism with a4; + a2 = a34 + a3 ina folded or an uncertainty configuration. This is because for any input angular speed w4, the output crank aj2 can move in a clockwise or anticlockwise di-First-order instantaneous planar kinematics 110 Figure 3.19 A 4R mechanism in an uncertainty configuration. Exercise Figure 3.4(1a) rection about 7; relative to the reference frame. All four rotors now belong to a two-system, and fi}-flefrbsfihs on At this configuration, the system is redundant, for if the input velocity is spec- ified it is not possible to determine @3, w, and @. In addition, (3.72) can be expressed in the form —Xx4 =x -x, —u]|° of ]--[7 : | |. (3.73) oy, The solution of (3.73) is not unique. It is not possible to solve uniquely for @3, @, and w, because the 2 X 3 matrix is noninvertible. EXERCISE 3.4 1. The input crank of the 4R mechanism is undergoing an anticlockwise rotation of 10 rads/sec (see Exr. Fig. 3.4(1a)).First-order kinematics of a closed-loop 4R mechanism = 111 coupler Exercise Figure 3.4(1b) 13, o2¢_h_coupler a4 10 rad/sec Exercise Figure 3.4(1c) (a) Compute the angular velocities «1, «2, and @. (Again it is easier to solve three simultaneous equations for 1, a, and 3.) (b) Compute the instant motion of the coupler (i.e., the coordinates for the instant center of rotation and its angular velocity) by disconnecting the mech- anism at the third joint and by considering the coupler to be the end effector of the 2R manipulator (see Exr. Fig. 3.4(1b)). (ce) Repeat part b by disconnecting the mechanism at the second joint (see Exr. Fig. 3.4(1c)). (d) Compare the results obtained in parts (b) and (c) and draw the in- stant center on your figures.First-order instantaneous planar kinematics 112 3.15 Instantaneous kinematics of a 3R serial manipulator using displacement equations We study the instantaneous kinematics of serial manipulators by in- troducing the concepts of rotors (instantaneous rotations) and instantaneous translations. This method introduces lines and line segments at the outset, es- tablishes a geometrical meaning, and, hence, a firm geometrical foundation for the development. An alternative and more cumbersome approach that is followed in many texts on robotics is to form time derivatives of the displacement equations for the three degrees of freedom serial manipulators derived in Chapter 1. This method is included for the sake of completeness. It is not recommended by the author, and the reader will undoubtably appreciate the complexity and lack of geometric meaning of the development. By way of example, the in- stantaneous kinematics for the 3R serial manipulator illustrated in Figure 3.20 will be determined by taking time derivatives of the displacement equations. These equations (1.28-1.30) were derived previously and will simply be re- stated here; the coordinates for a point Q in the end effector together with the orientation of the end effector are given by XQ = ay2C1 + a3c1+ 2 + asgCi+ 2+ 35 (3.74) YQ = 41281 + 43514 2 + 434814 24 35 (3.75) Y= 0 + 02+ 6. (3.76) The first derivatives of (3.74)-(3.76) can be expressed in the form _.. bxe axg axg : vox (=xXQ) = 20,0 + 30” ote 30,0 (3.77) - aye avg ay Yay (=Yo) = 36! + 30 + 50,8 (3.78) = = 87, 4 2, + 2 =p + : 7 ow (=9) 30," 302 30,73 (3.79) where do dO = 4hInstantaneous kinematics of a 3R serial manipulator 113 Figure 3.20 A 3R serial manipulator. and Ox, SO = —(ayasy + arssi2 + as4si+2+3)s 00, Ox, Ox, nee = — (4938142 + 4348142+3)> oe = 434514243; a Se = Caines + ansciv2 + aac +2+3)s 00 a a on = (@z3€1+2 + 434C1+2+3), on = 34014243 OVE O Ya LO Maas (3.80) 30, 0 363 Equations 3.77-3.79 can be expressed in the matrix form Vox . oO Voy | (=T*) = F*] w | = Py, (3.81) eo 03, where w is the instantaneous rotational speed of link a34 relative to the fixed frame; vg, and vg, are the X and Y components of the translational velocityFirst-order instantaneous planar kinematics 114 of a point Q on a link a34; @;, @2, and @; are the instantaneous joint angu- lar velocities. Also, 00, 8b 303 r=| 2% 80 a @.82) 00, 062 363 1 1 1 is known as the Jacobian matrix. This is clearly different from the previous formulation, T=sy, (eq. 3.46) where, from (3.48), for the reference point O chosen coincident with the first joint, the Jacobian matrix is O ye ys J=]0 -x —x3]. (3.83) Lee eee In (3.83), (x2, y2) and (x3, y3) are the coordinates of the points 2 and 3, and the columns of J are the coordinates of the joint rotor axes. A rotor with co- ordinates T= | voy o quantifies the instant motion of the end effector, where (vox, Voy) are the x and y components of the velocity of a point in the end effector coincident with a reference point O, which, for this case, is coincident with the first joint axis. In (3.81) rs YOx T*=| vo, o. quantifies the instant motion of the end effector using the components of the velocity of a point Q. However, in this formulation 7* does not represent theInstantaneous kinematics of a 3R serial manipulator 115 coordinates of a rotor, and the columns of J* do not represent the joint rotor axes coordinates. The two results are easy to compare by expressing the partial derivatives in (3.80) in terms of the coordinates of points 2, 3, and Q, and from (3.74), (3.75), and axg _ axg LD eee eae Ll ees 20; YO 50 Go- yw) 3 8s (vo — Ys), ay ay yo = =xg—- Seu 3, 3a, "2 aa "2 ag, 72? ie The substitution of (3.84) into (3.77) and (3.78) yields Vox = —Yow — (Yo — y2)@2 — Wo — y3)ws, (3.85) Yay = Xa + (xg — x2) + (xg — x3)a3. (3.86) Regroup the terms on the right sides of (3.85) and (3.86), substitute «, + @2 + w3 = o, and re-arrange to yield Vox + wyg = youn + y303, (3.87) Voy — Oxg = — x27 — x30. (3.88) The instantaneous center of rotation of the end effector is denoted by point G (see Fig. 3.21). Therefore, Vo = @X 1g, Vo = WX Fr, (3.89) and Vo — Vp = @ X (rg — 1) = wX Rg, (3.90) or Vo = Vo — @ X Rg = vg — ak X (xgi + yoi) = vg + wlygi — xgi). G91) The scalar components of this equation are Yor = Yor + yg, (3.92)First-order instantaneous planar kinematics 116 Figure 3.21 Motion of the end effector about instantaneous center G. and Voy = Vy — @xg. 3.93) The substitution of these results into the left side of (3.87) and (3.88) yields Vox = Y2W2 + y303, (3.94) Voy = —X2@2 — X30, (3.95) which essentially reduces the coordinates T* to T and the Jacobian J* to J.4 Dualities of parallel and serial devices 4.1 Introduction The statics of a parallel manipulator with three parallel RPR kine- matic chains and the first-order kinematics of a serial manipulator with three revolute joints were studied in Chapters 2 and 3. The force equation for the three connectors of a parallel manipulator and the equation for the instanta- neous motion of the end effector of a serial manipulator are, respectively, w = ja, (eq. 2.67) and Posy. (eq. 3.46) The vectors A and y are 3 X 1 column matrices of the magnitudes of the con- nector forces A fh 5 and the joint angular speeds eo w |, ws. and j, J are, respectively, 3 X 3 matrices, whose columns are the line coor- dinates of the connectors and joint axes, respectively. 117Dualities of parallel and serial devices 118 Correspondingly, the axis coordinates for an infinitesimal rotation of the end effector of a serial manipulator can be expressed in the form SD = J 88, (eq. 3.50) where the vector 60 is a 3 X 1 column matrix of the infinitesimal joint dis- placements, 50, 50 = | 5 |. 563, Dividing the left and right sides of this equation by a small time increment 6t and writing i= jim 22 er ieee T= jim, ot and = jim, o G@= 1, 2, 3) yields (3.46). The analogy, or rather the duality, between statics and instantaneous kine- matics (see Chapter 3) was stated as follows: In statics a directed line seg- ment represents the “rectilinear” concept of a force, whereas in kinematics a directed line segment represents the “circular” concept of a rotor. Clearly, (2.60) and (3.40) can be considered to be dual: w = fiSi + foo + fads T= aS, + 25) + 0353. It remains to perform a static analysis for a serial manipulator and dually to determine the instantaneous motion of a parallel manipulator. Before pro- ceeding with this, it is necessary to explain what is meant by the mutual mo- ment of a pair of lines and how this quantity is intimately related to the sta- tics and kinematics of a rigid lamina. 4.2 Mutual moment, instantaneous power, and instantaneous work Figure 4.1 illustrates a line $; in the xy plane with ray coordinates $= sie 4)Mutual moment, instantaneous power, and work 119 Figure 4.1 Mutual moment of two lines. Assume that there is a second line $, drawn through the point G perpen- dicular to the xy plane with axis coordinates $= | -xg |. (4.2) The mutual moment is defined as §7§ (=87s,). From (4.1) and (4.2), a Yo §, S = [c, s+ rf) | —x6 1 =n +r, — (esi — yori). (4.3) From Figure 4.1, (xes; — yoci) = rj and therefore TS =n. (4.4) The mutual moment for this pair of normalized lines, which are mutually per- pendicular, is their common perpendicular distance r;, The mutual moment can be considered to be an invariant in that the common perpendicular dis- tance between a given pair of lines is the same no matter how a coordinate system is chosen. The product sts (= n) is invariant under the group of Eu-Dualities of parallel and serial devices 120 Figure 4.2 Instantaneous power generated by a force. clidean motions. Also, if one applies a force of magnitude fj on either line, then the moment about the other is fir. Assume now a force of magnitude f; on the line $; which acts upon a lam- ina that is modeled in Figure 4.2 by a rigid link r; connected to the ground via a revolute Jojnt at G, the link rotating with an instantaneous angular ve- locity w = im a The instantaneous power generated by the force is given by WT T= G87) (oS) = fiwsS = fori, (45) and, analogously, the instantaneous work is given by W™SD = (F37)(598) = fi54578 = fidGn. (46) These simple results demonstrate that not only are the quantities instanta- neous power and instantaneous work invariant with respect to the Euclidean group of motions, which they must be, but they are also intimately related to the geometry of lines. It is important to note that when r; = 0, the mutual moment is zero and the instantaneous power and instantaneous work are also zero. This means that if the line of action of the force intersects the axis of rotation then the force cannot produce motion no matter how great the intensity of the force, fi: Then the pair of lines is reciprocal to one another just as the force and rotor are reciprocal. Figure 4.3 illustrates a lamina which has a single twist of freedom aboutMutual moment, instantaneous power, and work 121 Figure 4.3 A rotor and a reciprocal two-system of forces. a rotor axis through G perpendicular to the xy plane, the coordinates of which are 0 T= 0] 0]. Il The system of forces reciprocal to this one-system rotor is a two-system. It is possible to select any pair of forces in the pencil, for example, forces with magnitudes f, and f, which act on lines through G with 6; = 0 and & = a2. Any force in the pencil with coordinates 1 0 w=f[O]+Al 1 0 0 is clearly reciprocal to the rotor because 0 WT =f. fs 01) 0 | =0. o. From a physical standpoint any force that acts upon the lamina and passes through G cannot produce motion.Dualities of parallel and serial devices 122 a ° ey TX Sa . tT, %? (1,0) f Figure 4.4 Two rotors and a reciprocal one-system. Consider now a two-system of twists 7; and T>, a serial 2R manipulator (see Fig. 4.4). Without loss of generality, the axes of the twists $; and $2 are located on the X axis. Their coordinates are, respectively, 0 0 @}O]} and a} -a |}. 1 1 The lamina has two degrees of freedom and the instant motion of the end ef- fector is any linear combination, 0 0 T= | 0 | + @ | —ay2 |. 1 1 Any resulting axis of rotation must pass through the X axis because : - p=] — at Terai T @na\2 (a + wn) w+ton) (w, + 2) 7 ee iLe., x = @4)2/(w@ + @), y = 0, Because there are two independent free- doms in the plane there must be a single force of constraint which cannot produce motion of the end effector. This is a force of any magnitude f whose line of action passes through the axes of the two rotors (see Fig. 4.4). From a physical standpoint, one can imagine that the end effector extends over theMutual moment, instantaneous power, and work 123 Figure 4.5 Instantaneous sliding motion. X axis. The force w cannot produce motion of the end effector regardless of the magnitude of f, The coordinates of the constraint force w are 1 w=slo 0 and wT F = 0. Assume now that a body instantaneously has a pure sliding motion. This can be modeled by a prismatic joint (see Fig. 4.5). The slider has a single freedom. The system of forces reciprocal to this one-system of freedom is a two-system which consists of any force w with coordinates W= fil si in the parallel pencil of forces perpendicular to the sliding motion T, which has coordinatesDualities of parallel and serial devices 124 q B Figure 4.6 An RP manipulator and its reciprocal one-system. together with a pure couple with coordinates 0 co} 0]. 1 Any linear combination Ci 0 w=fi)s;]+co] 0 i. iE cannot produce motion, and w and T are reciprocal because w7 T=0. Consider a two-system of twists, T; a rotor, and T> a pure translation. This can be modeled without any loss of generality by an RP manipulator (see Fig. 4.6). The coordinates of T; and T> are, respectively, 0 1 T= a] 0] and 7 =v. 0]. 1 0 The lamina has two independent freedoms and the instant motion of the end- effector is a linear combination, 0 1 T=a,|0]+v] 0]. 1 0Mutual moment, instantaneous power, and work 125 " TAY ir Figure 4.7 Displacements T;, Ta, and a reciprocal one-system. There must be a single force, a one-system reciprocal to this two-system. This must be the force in the planar pencil of parallel forces reciprocal to T which passes through the rotor axis of 7), the coordinates for which are 0 wef} i. 0 The reciprocity condition WT T = Ois satisfied, From a physical standpoint, one can imagine that the end effector extends over the y axis. So the force w cannot produce motion of the end effector regardless of the magnitude of f. Finally, consider a two-system that consists of pure translations T, and T>. This can be modeled by a PP manipulator (see Fig. 4.7). The coordinates for T; and T> are, respectively, 1 Ty=v,] 0] and t=] 1 0 0 The instant motion of the end effector is a pure translation and 1 0 T=v,)0]+] 1 0 0 There is a one-system, a pure couple with coordinates 0 w=cl]0], 1Dualities of parallel and serial devices 126 Figure 4.8 A pair of skew lines. which is reciprocal to this nonrotating assemblage and Ww? T= 0. From a physical standpoint, it is clear that a pair of equal and opposite forces (see Fig. 4.7) applied to the end effector can produce no motion. In summary, for the motion of a lamina in a plane, the dimension of the space of twists of freedom (dim T) and the dimension of the space of reci- procal forces (dim w), which are commonly defined as forces of constraint, add up to three, dim T + dim w = 3. When the lamina has three independent freedoms dim T = 3 and thus dim w = 0, ie., there are no constraints acting upon the lamina. For a lamina with two independent freedoms, dim T = 2, dim w = 1, and thus there is a single constraint, a single force reciprocal to the two-system. For a lamina with one freedom, dim T = 1, dim w = 2, and there is a two-system of constraints or a two-system of forces reciprocal to the one freedom. When there is no motion, the lamina is completely con- strained, dim w = 3 and dim T = 0. Finally, it is interesting to note that the mutual moment of a pair of skew lines $ and $; is given by S18 = S73, = —r; sin a, (47) Here (see Fig. 4.8) r; and @ are, respectively, the common perpendicular and the twist angle between the lines. If the Pliicker coordinates for the lines $; and $ are S; = [Lis Mi, Nis Pir Qe Ril” and S = [P, Q, R; L, M, NJ", then STS = ST 3, = LP + MQ + NR + LP; + MQ; + NR; (48)Mutual moment, instantaneous power, and work 127 Exercise Figure 4.1(1) EXERCISE 4.1 1. (a) Compute the mutual moment of the line $, in the xy plane and the line $ perpendicular to the xy plane through the point (1, 1) (see Exr. Fig. 4.1(1)). (b) Compute the mutual moment of the two lines when 0 is located at the points (1, 0) and (0, 1). (c) The origin of the coordinate system is translated to the point (1, 2) and the coordinate system is rotated 60 degrees anticlockwise. Determine the Plucker coordinates of the two lines in the new coordinate system and compute the mutual moment. 2. (a) The wheel that is rotating anticlockwise (see Exr. Fig. 4.1(2)) at an angular speed of 10 rads/sec. experiences an impulsive force of 5 Ibf. Compute the instant power gained by the system. Repeat the calculation for a clockwise rotation of 10 rads/sec. and comment on your results. (b) The two-system of constraints can be quantified by the base forces with magnitudes f; and fp acting along the X and Y axes as shown in the figure. Show that the impulsive force cannot be expressed as a linear combination of the base forces. 3. Compute the instant motion of the end effector of the 2R manipulator (i.e., locate the instant center and compute the angular speed) when w) = 2 rads/sec.,Dualities of parallel and serial devices 128 Exercise Figure 4.1(2) o| 4e & x Exercise Figure 4.1(3) @) = 8 rads/sec.; w; = —2 rads/sec., w, = —8 rads/sec.; w; = —2 rads/sec., @2 = 8 rads/sec.; w) = —8 rads/sec. w, = 2 rads/sec. that is, show that the in- stant center must lie on the x axis and that any force along the x axis is recipro- cal to the end-effector motion. Use Exercise Figure 4.1(3). 4. (a) The end effector maintains a point contact at P such that it is con- strained to move along the X axis, as illustrated in Exercise Figure 4.1(4a). De- termine dim T and dim w, and write a set of coordinates for the bases of the twistMutual moment, instantaneous power, and work 129 _ oO P Exercise Figure 4.1(4a) ¥ A (0, a) < 0.0) Exercise Figure 4.1(4b) space and the corresponding reciprocal space. Show that the coordinates you have chosen satisfy reciprocity. (b) The beam maintains contact with the Y and the X axes at points A and B, as shown in Exercise Figure 4.1(4b). Determine dim T and dim w, and write a set of coordinates for the bases of the twist space and the corresponding reciprocal space. Show that the coordinates you have chosen satisfy reciprocity. (c) The beam maintains contact at points A, B, and D (see Exr. Fig. 4.1(4c)). Determine dim T and dim w, and give a set of coordinates for the bases of the twist space and the corresponding reciprocal space. 4.3 A static analysis of a planar serial manipulator Consider a force with magnitude f that acts upon the end effector of a planar 3R manipulator on a line $ with ray coordinates §. The axis coordi- nates of the lines of the rotors $1, $2, and $3 are $1, $3, and $3. This forceDualities of parallel and serial devices 130 ry "TL pase A (0.4) oO eo a Exercise Figure 4.1(4c) will induce a resultant torque 7; at each joint, and at each joint the motor must generate an equilibrating torque —7; to maintain equilibrium (see Fig. 4.9). The resultant torques due to the applied force f alone (i.e., disregarding any effects of either gravity or the inertia of the moving parts) are given by a =fn, 12 = fr, 13 = fr3. (4.9) Now ry, r2 and r3 are the mutual moments of the rotor axes and the line $. Therefore a = f Sts = Sh, n= 5 553 = Si, 13 =f Shs = Shi. (4.10) Equation 4.10 can be expressed in matrix form as r=I™, (4.11) where 1 v1 SaaS. t=|m]| and J=|—-xy -x. —x3 |. (see (3.47) 3. 1 1 1Static analysis of a planar serial manipulator 131 Figure 4.9 Static analysis of a 3F serial manipulator. When a force is applied to the end effector it is a simple matter to com- pute the magnitudes of the resultant torque at each joint, and, subsequently, the equilibrants —1), —7, and —73. This is called reverse static analysis. Conversely, when the actual torques generated by the motors are known we need to compute the actual force applied to the end effector. This is called forward static analysis. The joint torque/force relationship, equation 4.11, can also be derived us- ing instant power. Assume that the end effector is moving instantaneously on a rotor with coordinates 7 and that the corresponding joint angular ve- locities are «}, «2, and w3. Equate the instant power generated at the joints @ 7; + @27) + 373, or in matrix form 7 7, to the instant power generated at the end effector 77 w, which yields y7r = 77¥. Furthermore, from (3.46), TT = y"J". Substituting this result into the right side of y" = 77 and pre- multiplying both sides by (y7)~! yields (4.11). Further, from (4.11), the coordinates for the force for a forward static analysis are given by w= (J), (4.12) The geometrical meaning of J~! (referred to in Chapter 3) was shown to beDualities of parallel and serial devices 132 47353, J =| azst, | /det J, (eq. 3.53) eT 128} where 51, 523, and $3; are the coordinates of the lines $12, $23, and $3) join- ing the pivots 1-2, 2-3, and 3-1, respectively, as shown in Figure 4.9. Now gr |r an385, Mrt=—uyr= 3185) J det J = [423823 431831 a12512] / det J. ays) Thus from (4.12): 1 W = [a23823 a31831 a12812] | 7 | / det J. (4.13) 73, We also show (in Chapter 3) that det J can be expressed in three ways: det J = ay36%, $1, (eq. 3.55) det J = ay5%, 3, (eq. 3.56) det J = ar2sT, $3. (eq. 3.57) The geometrical meaning of det J is now apparent. Equations 3.55, 3.56, and 3.57 are alternative expressions for twice the area of the triangle formed by the three turning joints. For example, (3.55) is the product of the side a3 by the perpendicular distance sts) from the first joint to side a3. Assume now that the end effector is in contact with a rigid body. Indi- vidually applied motor torques n|} fo 0 O},,[ mm]. and | 0 0 0 3. cause the end effector to apply contact forces to the body. The coordinates of these forces can be obtained by substituting, in turn, these motor torquesStatic analysis of a planar serial manipulator 133 into the right side of (4.13). After this procedure, the introduction of the cor- responding expression (3.55), (3.56), and (3.57) for det J yields 3 = Saaml5581 = fades (4.14) W= va = Sar/54 S2 = fads, (4.15) 0 = te = Siom/8,8s = firsie, (4.16) where 1 Soe 4.17 ES fs (4.17) 7 Be = fy, 4.18 2S fu (4.18) ae =f. (4.19) 51953 Hence, individually applied motor torques 7, 72, and 73 cause the end ef- fector to apply forces to the body along the lines $3, $31, and $12, respec- tively. Consider again the reverse kinematic solution for the twist equation for the 3R manipulator, T= o S, + wy S$, + 03 $3. (eq. 3.40) Solutions for w, @2, and 3 can be obtained directly from (3.40) by form- ing the reciprocal products of the left and right sides with the coordinates 573, 531, and 4 of the lines $23, $31, and $12, which yields, respectively, = ut 3.59. w= aS (eq. 3.59) tf a = Ble, (¢q, 3.60) $3, S2 iT oy = (eq. 3.61) = =p, Shy S3Dualities of parallel and serial devices 134 In forming the reciprocal product of (3.40) with 523, the products SS and 85 vanish. The line $3 intersects both the lines $2 and $s (see Fig. 4.9). Analogously, 8, and 383 vanish, and TS, and 5,8, vanish, because the line $3; instersects both the lines $, and $2, and the line $j2 intersects both the lines $; and $3. EXERCISE 4.2 1. Exercise Figure 4.2(1) illustrates two positions of a 3R manipulator. All the linear dimensions are in feet. (a) A force of 5 Ibf acts on the end effector as shown in the figure. Compute the corresponding resultant torques for the two positions and the equi- librating joint torques. (b) Assume that the end effector is in contact with some object and is in equilibrium with equilibrating joint torques 7) = 2, ~ = 3, and 73 = 4 (bf = ft). Compute the corresponding external forces that act upon the end effector for the two positions. 2. The end effector of the 3R manipulator is instantaneously rotating about point C(O, 2), as illustrated in Exercise Figure 4.2(2). The velocity of point P is vp = 1 in/sec. Determine the angular velocity of the lamina and the velocity vo of a point in the lamina coincident with the reference point O. Compute the corre- sponding values for the joint speeds w, «2, and w; from the twist equation T= o8, + on8; + 0353 by forming reciprocal products with the coordinates of lines $23, $31, and $12, re- spectively. Compare your results with part 1 of Exercise 3.2. 3. We require that the end effector of the RPR manipulator rotate on an instant center with coordinates (2, 1) with a clockwise angular speed of 10 rads/sec. Compute @), v2, and @; by forming in turn reciprocal products of the twist equa- tion with three lines; each line is reciprocal to pairs of joint motions, as shown in Exercise Figure 4.2(3). 4. We show for a 3R manipulator that det J = ayo5y, T Ss (see (3.57)), which is twice the area of the triangle 123 shown in Exercise Figure 4.2(4). (a) Deduce that det J = a12 a3 sin 6). (bb) Deduce that the maximum value for det J is det Jmnax = @12023 and A = det J/ det Jmax = sin 0. (©) Sketch the graph A = sin @) for 0 = 6 = 360 degrees. Draw the pair configurations of the manipulator for A = 0 and A= + 1. (Note that the ra- tio A is independent of the manipulator’s dimensions.)Static analysis of a planar serial manipulator 135 Y S lof 5 lbf a ‘ ©, 2.5) G4 ase fu ia (X35 Ys) a5 = 1.5 ft (X25 Y2) Vy, a2 = 2ft 8, Exercise Figure 4.2(2) 3 [99 2.5.15) Exercise Figure 4.2(3)Dualities of parallel and serial devices 136 Exercise Figure 4.2(4) 4.4 Asstatic analysis of a parallel manipulator with RPR and 3R kinematic chains The skeletal forms of planar motion parallel manipulators with RPR and 3R serial chains are illustrated in Figures 4.10 and 4.11. Before pro- ceeding with a static analysis of a parallel manipulator with 3R serial chains it is interesting to restate equations 2.88-2.90 for the reverse static analysis of a manipulator with three RPR serial chains, viz., KS, f= Tino, f= Tyas s., which were obtained by inverting the j matrix. These equations can be ob- tained directly from (2.60), w= fil + faa + fads, by forming, in turn, the reciprocal products of the left and right sides of (2.60) with the coordinates $53, $3), and $1 of the lines $53, $31, and $1 (see Fig- ure 4.10). It is now possible to perform a static analysis of a parallel manipulator with three 3R serial chains using the results of Section 4.3. Admittedly, the notation can become tedious because it is desirable to label the sequence of joints in a serial chain by 1, 2, 3, beginning with the grounded joint. Fur- thermore, it is also desirable to label each of the three serial chains by 1, 2, 3. This is accomplished in Fig. 4.11 by introducing subscripts. Assume that we actuate the device by introducing motors at the joints (1),Static analysis of a parallel manipulator 137 Figure 4.10 Parallel device with RPR serial chains. Figure 4.11 Parallel device with 3A serial chains. (1)2, and (1)3 and that torques (71)1, (71)2, and (7;)3 are applied. Then from (4.14) and (4.17) the coordinates of the resultant force acting upon the mov- able lamina are given by W = (W231 + (v23)2 + (o023)3, (4.20) or W = (fa Gas) + (fas)2 (S23)2 + (fas)3 23), (4.21)Dualities of parallel and serial devices 138 where (m1 (m2 (m3 eer eee eee ee =r? (4.22) f— Ey Gor =" Een’ "Eye, 6 and (4.21) can be expressed in matrix form by w = [(23)1, G2a)2, (S23)s] 3, (4.23) where (fas fos = | (fase |- (fa3)3 Let us actuate the parallel manipulator by introducing motors at joints (2)), (2)o, and (2)3, and torques (72);, (72)2, and (7))3 are applied. Then from (4.18) and (4.15) the coordinates of the resultant force that act upon the movable lamina are given by W = (Wai) + (War)2 + (War)s, (4.24) or w= (fairnGat + GaroGav2 + FaraGan)a, (4.25) where Bo Bre Oo Baye Oo Bey 420 and (4.25) can be expressed in matrix form as w= (Gai, Garda, Gai)s] far, (4.27) where fn far = | Garde |- (fais. Analogously, if joints (3)), (3)2, and (3)3 are actuated, the resultant force that acts upon the movable lamina is w= [(612)1, (612)2, G12)3] fia, (4.28)Static analysis of a parallel manipulator 139 10 lof 10 lef Cr C2, Cy fs 120° . BX 10°" By, By Exercise Figure 4.3(1a) where (fia) fio = | fia)2 |. (fi2)3. EXERCISE 4.3 fe (a) Determine the unitized coordinates of the lines $;, $2, and $3 that join the connectors ByC}, ByC2, and B3C3. Use point By(B2) as your reference (see also part 1 of Exercise 2.3). Then, determine the coordinates of the points of intersection of the pairs of lines ($2, $3), ($3, $1), and ($1, $2). Hence deter- mine the coordinates of the lines $23, $31, and $2 which are, respectively, reci- procal to the line pairs ($2, $3), ($3, $1), and ($1, $2). Following this, obtain ex- pressions for the resultants fi, 5, and f; from the force equation w= isi + fbr + fads. Use these expressions to compute the resultants for a vertical force of 10 lbf acting through point C; (see Exr. Fig. 4.3(1a)). Following this, compute the resultants for a vertical force of 10 lbf acting through point C2 (C3), as shown. (a) Repeat the previous exercise for the force of 5 Ibf that acts upon the truss (see Exr. Fig. 4.3(1b)). Following this, compute the resultants when the force of 5 Ibf acts vertically through point C, and then through point C2 (C3).Dualities of parallel and serial devices 140 Exercise Figure 4.3(1b) (3); (3), a 135° (2); (2), 25" (1), P80” , Exercise Figure 4.3(2) 2. The upper platform of the drawing in Exercise Figure 4.3(2) is in contact equi- librium with another body. Anticlockwise unit joint torques of 1 Ibf inch are ap- plied at the three base joints, the three intermediate joints, and the three upper platform joints. Compute the corresponding external forces acting upon the up- per platform for each of these three cases. 4.5 A kinematic analysis of a parallel manipulator Figure 4.12 illustrates a movable lamina connected to a fixed base via three in-parallel RPR kinematic chains. Assume that the prismatic pair in each chain is actuated, and that the moving platform undergoes an instanta- neous rotation about some point G, which is called the instantaneous centerKinematic analysis of a parallel manipulator 141 Figure 4.12 A parallel! manipulator. for the rotation. This can be modeled by a revolute joint at G (connected to the ground) to which the movable lamina is rigidly connected. The coordinates for the instantaneous motion are known and measured relative to some reference point 0 in the fixed base. ‘We need to compute the sliding velocities v1, v2, and v3 of the actuated pris- matic pairs. Assume that the geometry of the system is known and the co- ordinates of the lines $1, $2, and $3 are, respectively, $1, $2, and 53. Consider the instantaneous kinematics of any one of the RPR chains (see Fig. 4.13), which is labeled with the subscript i, that connects the platform to the base triangle. The instantaneous first-order kinematics of the platform is given by T= T+ Tit Ti, (4.29) which can be expressed in the form T= oy Si + Fa t+ oS, (4.30) where «4, and ws; are the angular joint speeds and $1; and 53; are the axis co- ordinates for the revolute joints, and 73; = [v,, vy; OJ" = vile, 5% OJ" are the coordinates for the slider motion. Compute the slider velocity v;, which is the velocity component of the moving pivot C; parallel to $;. This is easy to accomplish with the reciproc-Dualities of parallel and serial devices 142 ond Figure 4.13 The instantaneous first-order kinematics of an RPR chain. ity condition developed in Section 4.2. Clearly, line $; intersects the axes of both revolute joints. Therefore S781, = 5785 = 0. (4.31) Now §; = [ci si pil”, and hence ci $7 Ta; = vi leis 533 P| Si 0 = vic} + $2) = vi. (4.32) The formation of the reciprocal product of the left and right sides of (4.30) with 3; and the substitution of (4.31) and (4.32) yields v= st T (4.33) Hence, for the three connecting RPR chains, y= st, v= 597, v3 = ST 7, (4.34)Kinematic analysis of a parallel manipulator 143 which can be expressed in the matrix form v=jTT, (4.35) where VI 1 C2 3 ve=|vo|andj=|s1 so 53 |. (see (2.68) V3, 1 P2 P3. When the instantaneous twist of the platform is specified, it is a simple matter to compute the values for the actuator speeds vj, v2, and v3. This is called the reverse or inverse velocity analysis. Conversely, when the actua- tor speeds are known, then we need to compute the actual instantaneous twist of the platform. This is called the forward velocity analysis. From (4.35): =v. (4.36) Equation 4.36 can also be derived using instant power. Assume that a force with coordinates w is acting instantaneously on the moving platform and the corresponding forces generated in the connectors are fi, fo, and f;. Equate the instant power generated in the joints ATv(AT = [fi,. fy fs) to the instant power generated by the platform w? 7 to yield ATv = 7 7. Substituting w7 = AZT (see (2.67)) and premultiplying both sides by (A7)~! yields (4.35). Chapter 2 shows that the geometric meaning of j~! is S71 = [3-28], 1-389, s1 297 1/ det j, (eq. 2.82) where $12, Sos, and Ss are the coordinates of the lines $12, $23, and $3, pass- ing through the points of intersection of the connector lines $, $2, and $3, as illustrated in Figure 4.10. Now GY = GYD = [93-28%, 51-384, 52-187 IT det j = [83-2823 51-3831 52-1512] / det j and thus from (4.36): vy = [53-2523 51-3831 $2—-1Si2] | ve | / det j. (4.37) V3,Dualities of parallel and serial devices 144 We also show in Chapter 2 that det j can be expressed in three ways: det j = s3-283,51, (eq. 2.84) det j = 51-387,52, (eq. 2.85) det j = 5.—187,53. (eq. 2.86) The geometrical meaning of det j is now apparent in the sense that it can be expressed to a scalar multiple as any one of the reciprocal products of the pairs of lines ($23, $1), ($31, $2), and ($12, $3). It is also useful to compute the component v,; of the velocity of point C; that is perpendicular to the line $;. The coordinates of the line $;, are =5) Sic =] Ct & This line intersects the axis of the revolute joint at C;, and it is also recipro- cal to the sliding motion ci Tr = Vi} Si 0 because Ci Ho Tai = ThSic = vi (-Si cis Ci) | 8 | = 0. 0 Finally, the reciprocal product 0 ST. Ty = con (—8; 4 Ci) | 0 | = yl = vu 1 Hence, the formation of the reciprocal product of the left and right sides ofKinematic analysis of a parallel manipulator 145 (4.30) with §, yields v,j = are given by 7c F, and therefore the velocities vq, v2, and vz (4.38) which can be expressed in the matrix form w= (crf, (4.39) where v, = [Yas Yea Ya!” and [C] = [Sic $2c 3c). ‘We may now deduce the corresponding results for an instantaneous rota- tion 5¢ of the moving platform about G. From (4.35): 8) 8, | =j? ob, (4.40) 5e3, which is the expression for the vector of the displacements of the moving pivots C; parallel to the lines $;. Clearly, 8€,/8t e v= Jim | ¢o/6r] and #= jim, 2 5€3/61. From (4.39) the vector of the tangential displacements of the moving pivots C; is given by €; 66, : € 86) | = (CISD. (441) €3 503 Thus, €; 56/51 ve = jim, | €2 5602/5 €3 563/51 It is instructive to deduce (4.40) and (4.41) from first principles, becauseDualities of parallel and serial devices 146 Table 4.1 The dualities of parallel and serial manipulators Parallel Serial Wrench coordinates Twist coordinates w= (Fo) T= (Vo; @) 8D = (Xo, yo; 5b) Twist coordinates Wrench coordinates T= (Vo, @) w= (f; Co) 8D = (8x0, yo, 84) Scalar connector forces Scalar joint velocities A= ef) ¥ = (@, @2, @3) Scalar joint displacements 50 = (501, 56), 583) Forward statics Forward kinematics wHjr T=Jy 8D = J60 Connector velocities v Joint torques + Connector deflections (5€), 5€, 5€3) Inverse kinematics Inverse statics v=j'T r= I 0 [5€), 5€2, 8€3]” these results are important in the analysis of a compliant parallel manipula- tor (see Chapter 5). Before we proceed with this analysis, the instantaneous kinematics of parallel manipulators in singular configurations are analyzed and the dualities between serial and parallel manipulators are listed in Table 4.1. 4.6 Instantaneous kinematics of parallel manipulators in singularity configurations We studied the statics of parallel manipulators in Section 2.9, which showed that a singularity condition occurs when the connector lines meet in a finite point Q. In addition, the three connector forces belong to a pencil of forces that pass through @Q and are linearly dependent (see Fig. 4.14).Instantaneous kinematics of parallel manipulators 147 B, B; Figure 4.14 Singularity condition; the connector lines meet at Q. In this configuration, the moving platform has an uncontrollable instant mobility and instantaneous angular velocity @ # O about Q. Because of this, points C; (i = 1, 2, 3) have velocities v,, = wQC; perpendicular to the con- nector lines, and hence, the actuator sliding velocities v; are zero. This is easy to show by choosing @Q as the origin of an xy coordinate sys- tem for which cr 2&3 J=] 51 S283]. 000 0 v=|0 0 yields 0 ca 8 0 Ya Ol=]e2 s2 0]! —x6], (4.42) 0 cz 53° 0 1Dualities of parallel and serial devices 148 where (xg, yg) are the coordinates of the instant center. Expand (4.42) to get 0 = @ {yeti — x65 + 0}, 0 = @ {yger — xG82 + 0}, 0 = {yGc3 — xG53 + 0}. (4.43) For w # 0 the only solution of set (4.43) is xg = yg = because generally S40, [1 N40, [2 Sl 40, en C2 cz 83 S253 i.e., any pair of connector lines are assumed not to be collinear. Hence, the instant center is Q. As Q moves toward infinity the lines become parallel and the instant center G moves toward infinity, and the instantaneous mobility is an instantaneous translational velocity in a direction perpendicular to the con- nector lines (see also Fig. 2.27 where this velocity is parallel to the y axis). 4.7 An infinitesimal displacement analysis for a parallel manipulator Figure 4.15 illustrates the moving platform of a parallel manipulator undergoing an infinitesimal rotation 5¢ about an axis $ through a point G. This is modeled by a revolute joint at G (connected to the frame of reference in the fixed body) to which the moving platform is rigidly connected. The lines $13 and $j are perpendicular to the lines of the connectors $; (i = 1, 2, 3), and they pass through the fixed and moving pivots B; and C;, respectively. When the moving platform rotates relative to the fixed platform about an axis through G, each moving pivot C; displaces to a point C;. This displace- ment can be decomposed into two displacements, an infinitesimal displace ment 6€; along the line $; together with an infinitesimal displacement €,66;, which is tangent to a circle of radius €; centered at B; (see Fig. 4.16). It should be clear from this figure that 8; = 5p and €,56; = ricdh, (4.44) where r; and ric are, respectively, the perpendicular distances from G to the lines $; and $;c. Therefore n= 578, and nc = 57.8, (4.45)Infinitesimal displacement analysis 149 Figure 4.16 Infinitesimal displacement of a single connector.Dualities of parallel and serial devices 150 where §;, Sic, and § are the coordinates of the lines $;, $;c, and the axis $ through G. The substitution of (4.45) into (4.44) yields 8€; = SK5bS) = STD and €,80; = (4.46) Hence, for i = 1, 2, 3, 8e, = sTED, 82 = SED, 83 = ST8D, (4.47) which can be expressed in matrix form as ae) 8l> | = jTD. (4.48) Bes, This equation is precisely (4.40). Also, for i = 1, 2, 3, €150, = 81 8D, 6586) = 57 6D, €3603 = STD, (4.49) which can be expressed in matrix form as €150; €50, | = [C]"8D. (4.50) £3803, This equation is precisely (4.41). Choose a reference point O in the fixed platform to determine the coor- dinates of the lines $;, $18, and $c (see Fig. 4.17). The matrices for the coordinates of lines $;g and $;c, which are perpendic- ular to line $; and pass through the fixed and moving pivots are, respectively, Peete [B=] cr co cg |, (4.51) qs QB 3BDifferential of a line 151 [l=] a co |, (4.52) jc Re Bec where gic = gig + €;. Observe from Figures 4.15 and 4.16 that when the moving platform ro- tates through an angle 5¢, each connector moves from a line $; to a new line $7 due to a small rotation 66; about the fixed pivot B;. We can now deter- mine the coordinates for $*. 4.8 The differential of a line Assume that the line $; drawn in Figure 4.18 is attached to a revo- lute joint located at point B; and is connected to the ground (the page), which is the frame of reference. For a small rotation 56,, line $; moves to a second line $7 and every point on $; undergoes a small rotation, except for the point that is coincident with B;. Consider that the coordinates 3; of line $; are para- metric functions of 6, and $1 = 5 (6). (4.53)Dualities of parallel and serial devices 152 Figure 4.18 The differential of a line. Therefore, the differential of the coordinates is 85; = a 56;. (4.54) From Figure 4.18, the coordinates of $, are $1 = (ci 58 pi} = (ci, 5;; OB sin (6; — a)}. (4.55) Write Sig = d5;/d0; to yield a ds; Sip = aa = (—5a cx OB cos (0, — «)} = (-5:. ci Gis} (4.56) i These are the coordinates for line $;s, which is defined as the geometrical differential of $;. In addition, the coordinates of st are given by S* = 5; + 51380; (4.57) which are linear combinations of the coordinates of lines $; and $j.5 The stiffness mapping for a parallel manipulator 5.1 A derivation of the stiffness mapping Figure 5.1 illustrates an elastically compliant, planar parallel ma- nipulator. The moving and fixed platforms are connected by three RPR ser- ial chains and in each prismatic pair there is a linear spring. Assume that the moving platform is in equilibrium with an externally applied force with co- ordinates W and magnitude f is applied to it on a line $. Then W= fiSi + frSo + Abs, (eq. 2.60) where f;, fo, and f; are the magnitudes of the resultant forces in the connec- tors and $1, §2, and $3 are the line coordinates of the connectors. A small change w in the applied force will cause the upper platform to move with an infinitesimal rotation with coordinates 6D on an axis perpen- dicular to the page through a point G, as illustrated by Figure 5.2. These quan- tities are related by a3 X 3 stiffness matrix [K] which we will determine next. Assume that the free lengths of the springs ¢,; and the stiffness constants k, are known, together with the coordinates w of the applied force and the coordinates $1, 2, and S3 of the lines $1, $2, and $3. Assume also that the movable lamina is initially loaded and the spring lengths are ¢), €>, and €3. Then substitute the relationships f; = k; (€; — foi) (i = 1, 2, 3) into (2.60), where k; are the spring constants and (¢; — €,;) is the difference between the current and free length of an ith spring, to yield w = kr — Corsi + ka(€o — Lor)82 + ka(€3 — €oa)S3- 6.) 153The stiffness mapping for a parallel manipulator 154 Figure 5.1 A planar compliant coupling. The total derivative of (5.1) can be expressed in the form SW = ky SCS) + kySlrs2 + kxSl383 as) dn + = C1) S150, + = Cy) 2, ker fo 76, 50, + hallo (00) 19,0 = bg) 22 + ka(l3 fea) 79,08 = $1 ky 8) + $2 ky 5lq + $3 kg Sls + Fipki(1 — p)€1501 + Sapko(1 — p2)€2602 + S3—(1 — p3)€366, (5.2) where p; = €,;/€; and d§;/d0; = S;g are the coordinates of a line $j perpen- dicular to $; that passes through a fixed pivot B; (see Fig. 5.2 and equation (4.56)). Equation 5.2 can now be expressed in matrix form as 5, €:60, Sw = [81 $2 $3] [k] | 5€2 | + [81a $2 S35] [k(1 — p)] | €256 |, (5.3) ls £3583Derivation of the stiffness mapping 155 Figure 5.2 Motion of the platform with respect to a reference frame in the fixed platform. where [k] and [k(1 — p)] are the (3 X 3) diagonal matrices k 0 0 kx(1— pi) 0 0 [=| 0 k. Of and [kK -p]=]0 kl — pr) 0 0 0 & 0 0 1a(1 — ps) The following notation was employed in Chapters 2 and 4: c1 62 63 F=HiHHl=] 1 2 53 Pi P2 P3., is the matrix of the coordinates of the lines $1, $2, and $3, and —s1 82 —$3 (B] = [S18 Sop S38) =] c1 2 o3 1B 928 3BThe stiffness mapping for a parallel manipulator 156 is the matrix of the coordinates of the lines $1g, $28, and $3g (see Fig. 5.2). Also, from (4.48) and (4.50): bey £50, a 82 |=j7SD and | £250) | = [C]78D, 5e3, £3503 where Peel Neti: [Cl = [icS2cSacl=| cr 2s jc Re Be is the matrix of the coordinates of the lines $1, $2c, and $3¢ (see Fig. 5.2). Substituting these expressions, (5.3) can be expressed in the abbreviated form di = Yj [Kh J7 + (B) k 1 - pl (CI) 8D, (5.4) and, finally, (5.4) can be expressed as dw = [K] 5D, (5.5) where the required stiffness matrix [K] is given by (K] = jfk? + [B] [k (1 — p)IC]?. 6.6) Clearly, [K] is symmetrical only at the unloaded position for which p; = 1 or €; = 5; and [K] = j{klj". It is easy to deduce from (3.33) that under the action of the Euclidean group, a twist 5D” expressed in a new coordinate system is related to the same twist quantified in the original coordinate system by 8D = [E] 8D". (5.7) Similarly, an incremental change of force quantified in the new coordinate systems is related to the change of force quantified in the original coordinate system by (see (2.52)) dw = [e] 50", (5.8) and j= [els”. (5.9)Dimensions of the elements of the stiffness matrix 157 Analogously, [B] = [e] [B’] and [C] = [e][C"]. Substitute (5.7)-(5.9) into (5.4) to yield {e] 8” = [e] 47" (ky? + (B") (k A — p)] [C'} Le]? (E] 8D". (5.10) From (3.37), [e]” = [E]~!. Substituting this result in (5.10) and premultiply- ing the left and right sides by [e]~! yields 8%" = [K"] 8B", (5.11) where [K"] = {j” (k] j"7 + [B’] [k (1 — p)] [C"J"}. Equation 5.11 expresses the relationship between the twist produced by a change of force in a new coordinate system. Equations 5.4 and 5.11 are expressions that relate the same twist expressed in terms of the old and new coordinate systems produced by the same force increment expressed in the same old and new coordinate systems. It is also clear that the stiffness matrix [K] itself changes. Substitute (5.7) and (5.8) into (5.5) to give (e] 80” = [K] [E] 8D", or 8i0" = [e]"! [K] [E] 8D", (5.12) and it follows from (3.37) that [e]~! = [E]’. Therefore, the new stiffness ma- trix [K”] is related to the old matrix by (K"] = (EV (K1 (EI. (5.13) 5.2 The dimensions of the elements of the stiffness matrix It is interesting to examine the dimensions of the elements of the stiffness matrix 1K] = j (1 7 + (B] [KC — p)) (Cy. (eq. 5.6) The dimensions of the two terms on the right side of (5.6) are clearly iden- tical because j, [B], and [C] are matrices whose columns are the coordinates of lines in the xy plane. Furthermore, p is dimensionless. Without loss of gen- erality, the dimensions of the elements of the matrix [K] will be determinedThe stiffness mapping for a parallel manipulator 158 for the symmetric case by substituting p = 1 in (5.6) and expanding the right side, which yields ac otk O Offa ss pr [K]=] 51 s2 s3]|}0 kz Offer sp pp}. (5.14) 1 P2 P31LO 0 ksjLes s3 ps Therefore, crky coke caks|]cr spi (K] =| siky soko sks |] cr 52 pr |- (5.15) 11 Poke psksi Les $3. ps It follows that [K] is a 3 X 3 symmetric matrix which can be expressed in the form ku kar kar [K] =| kor kor kp J, (5.16) Ka kar ka, where expanding the right side of (5.15) and equating to the elements of (5.16) yields 3 k= x Cikis 3 3 kn = > cisiki, kao = > Sk; 3 3 3 ka =>) cwiki, kee = > sivik, kas = >. poke. (5.17) T T T It is important to recognize that the dimensions (dim) of the elements of [K] are not all the same: dim (k;) = force/length = FL, dim (p,) = length = L, and dim (¢;, 5;) = 1. Hence FL“! FL"! F dim [K]=] FL"! FL“! F |. (5.18) F fa FL.Stiffness mapping of a planar two-spring system 159 ky 6, B, x Figure 5.3 A two-spring system. Now : L dim 6D =| L]. (5.19) 1 Therefore, from (5.18) and (5.19): F dim ({K] 8D) =| F | = dim dw. (5.20) FL. The dimensions of (5.18) are thus consistent. 5.3 The stiffness mapping of a planar two-spring system Figure 5.3 is a schematic representation of a pair of RPR serial con- nectors, with springs in the prismatic joints, and connections to the ground at points B, and B>. The free ends are connected to a common turning joint at a point C. This system can be obtained from Figure 5.1 by shrinking the movable platform to a point C and removing the third RPR connector. Equation 5.5 can then be expressed by of = [K] 5D, (5.21) «lel whereThe stiffness mapping for a parallel manipulator 160 is a small increment of force acting at point C and »-[5| is a small displacement of point C. The stiffness matrix can be expressed in the form (K] = 7 (kK) j7 + [C] [kK — p)] (CV, (5.22) where all the submatrices are now 2 X 2, and it EB | (5.23) S152 sai ie (aS Letters. [B] = [Cc] [ ei | (5.24) Furthermore, _|h 0 [kK] = i; | (5.25) and ki — py) 0 kd — = 7 5.26) (k(1 — p)] f 0 bu - ook (5.26) It is left to the reader to deduce this result by writing the equation f=fisi + Aso, (5.27) where r=/*| and s=[°) s.=| Fy SL 52 Then introduce the spring constants ky and kz into (5.27), which yields £ = ki(€, — Corsi + ka(€2 — €o2)82. (5.28) Let us take a total derivative of (5.28) (see (5.2)) and make the necessary substitutions by simplifying the results from Chapter 4. Now assume that a wheel is connected to a platform using a two-springStiffness mapping of a planar two-spring system 161 Figure 5.4 A passive two-parameter spring actuated by a P-P manipulator. system, as illustrated in Figure 5.4. The platform is connected to the ground by a serial pair of actuated prismatic joints that are tuned for fine position control. The wheel maintains contact with a rigid wall. From (5.6), [K] is a symmetric 2 x 2 matrix and [K] = [Ko] + [Kp], (5.29) fer elf Offa tko1=[* alle ile A 6.30) _ [9 —s82] [ad — pd 0 —s1 Cy wi=| 2 ell 0 wt po|lon h fae where p; = €4:/€;. where andThe stiffness mapping for a parallel manipulator 162 Now assume that the system is close to its unloaded position (€; = €,;). From (5.29): cr col[ki Offa s = . 32 i [3 2| (* i [: “| oe Assume that the values 6; = 45 degrees, @, = 90 degrees, kj = ky = 10 Ibf/in. gives 5 wi=[3 | Ibffin. Assume initially that the center point C is completely constrained by clamping the wheel to the wall. This means that any external force f = fi + Jy j acting through C can be applied to the wheel. Also, any small displace- ment of the lamina B, Bo, relative to the slider displacements 6d and 8d, will cause a small change in f because of the changes in the lengths of the pair of springs. Now o-(elts and therefore of] __[5 5] [8a falo-Es elle] ~ Invert the matrix in (5.33) to give da] _ _[ 03 -0.1][ a | fe ls 0.1 iI Cae The subscript c denotes correcting slider displacements which control the contact force. Equation 5.34 can be used to control a time-varying contact force f = fai + Sf,j between the fully constrained wheel and its environment. At each instant, an error in force is known which can be reduced using (5.34) by computing the proper force error-reducing displacement (or force-correcting displace- ment) SD, = dey i + Sde j. Assume now that the wheel is at rest, but that it is loaded with an exces-Stiffness mapping of a planar two-spring system 163 sive normal force parallel to the vector u, (see Fig. 5.4). This force, which is reciprocal to the wheel motions, can thus be reduced without moving the wheel provided that _ [a] _ <, [cos 45] _ .,. [0.707 aan | = on B | = hn ee Ce where 6f,, is the desired change in the normal force. Substitute this result in (5.34) to give _[8de] _ ., [-0.1414in.Abf D. eal a%| o | (5.36) So, for this example, 6d, = —0.1414 df, and 8d,2 = 0. Therefore, a dis- placement in the negative x direction (for a positive 8f,,) reduces a compres- sive normal force that passes through C. A compressive contact force f, is negative and hence a positive 8f, reduces an excessively compressive con- tact force. It is important to recognize that the wheel does not move when the lamina B,B, moves in the x direction. Hence, this is inherently the best direction for correcting a normal force error. Intuitively, one may consider, at the outset, that a motion of the laminar parallel to u,, would change the contact force without moving the wheel. This is clearly not the case. Finally, assume that the wheel C moves upward along the wall. The dis- placement of the wheel center C is given by SD, = 8d, i + Sdz j, (5.37) where 8d, = dp, cos 135, 8d. = Sp, sin 135, and dp, is a small displacement of point C parallel to the wall. Here the subscript t denotes the displacement 8D, of the wheel, which is tangential to the normal contact force produced by the slider displacements, 8d,; and 5d. Hence 8D, = dp, (—707i + 0.7079), (5.38) and expressing (5.38) in matrix form yields —0.707 8D, = | eae | (5.39) It is important to recognize that 5D, is the only allowable freedom for point C, assuming that contact between the wheel and the wall is to be maintained.The stiffness mapping for a parallel manipulator 164 Also, a change in the constraint force 5f,, cannot produce motion. As stated earlier, such forces and displacements are reciprocal and 7 =0 atASD, = 4f [0.707, 0.707] a a It follows that the small slider displacements 5d, and 8d, can be used for the simultaneous control of normal force and tangential motion. By super- position dy) -0.1414 -0.707 [iat |= Ga o.oo] * 52] ort 6.40) In (5.40) G, and G, are dimensionless scalar gains, and Sp, and $f, are errors in the wheel position and the normal contact force. 5.4 Force and motion control using a serial manipulator with a compliant wrist Figure 5.5 illustrates a planar three-revolute serial manipulator with a compliant wrist. The workpiece, which is held fixed in the gripper, is in contact with a fixed rigid lamina at a single point P. Hence, there is a single constraint force that acts on the workpiece along the line $, and the work- piece thus has two freedoms: a pure translational displacement along the sur- face of the lamina and a rotation about the contact point P. Assume that the workpiece is to remain in contact with the rigid lamina. The compliant wrist is an in-parallel manipulator with special geometry (see Fig. 5.6). Coaxial turning joints B, and B are labeled By and the coax- ial tuning joints C2, C3 are labeled C23. The forward analysis for this device is given in subsection 1.6.3. Each of the three RPR connectors has linear springs. The revolute joints of the serial manipulator are actuated, and changes in the angular displacements produced by the actuators are denoted by (i, Sy, 5yf3). Assume the manipulator is tuned for fine position control so that it is nonback-drivable when serving. The three RPR connectors are not acu- tated in the application, and the wrist is thus passive. The relative location of the pair of platforms is controlled by actuating the joints of the serial ma- nipulator. Assume initially that the tool is fully constrained by clamping it to theForce and motion control using a serial manipulator 165 dy, oo) Figure 5.5 A passive three-parameter spring actuated by a 3R manipulator. ground. This means that any force with coordinates w applied to the work- piece by the ground does not move the platform or robot manipulator. How- ever, any twist of the movable lamina (B, — B3) with coordinates 8Dr mea- sured relative to the ground will change the force applied to the workpiece, and this can be expressed by dW = —[K]8Dp. (5.41) The negative sign was introduced because the twist of the platform (C; — C»3) relative to the grounded workpiece was specified previously. Assume that the contact force between the workpiece and the ground is to be reduced by an amount 5v. The coordinates of the required twist 5D can be computed from (5.41) and Dp = —[K\"} 50. (5.42)The stiffness mapping for a parallel manipulator 166 Br : B, Figure 5.6 A parallel device with special geometry. Following this, the required joint motions Sy, dy2, and Sy can be com- puted from a reverse analysis of the 3R manipulator, and using (3.52): oh : Sy | =I“! SDp. (5.43) Sts Suppose that the tool is in point contact with the ground at P (see Fig. 5.5). The line of action of the normal constraint force is labeled $,. We need to determine a point F about which the base movable lamina will rotate to control (and even reduce to zero, if required) the constraint or contact force at P. From (5.42): 8De = -[K]7! du, (5.44) where 812 = 5f[ca Sa: ral” and SDr = 8¢lyr, —xp; 1]". The tequired point of rotation F with coordinates (xp, yr) is determined from (5.44) by the stiff- ness mapping [K]. A rotation of the movable lamina about F cannot move the tool because the change of force applied to it is acting on the line $,. Point F is therefore the best point to use to control the contact force because a twist with coor- dinates 8D cannot cause motion of the tool. Assume finally that the base movable lamina undergoes an infinitesimal twist about an axis through some point G. This twist can be decomposed into a twist of the workpiece about an axis through point E, together with a twistForce and motion control using a serial manipulator 167 Foo Figure 5.7 Pencil of lines through point F. about an axis through point F (see Fig. 5.5) which simply alters the contact force along $,, and 8Dg = SDp + dDe- (5.45) The line $, that joins points F and G intersects $, at point E. The workpiece twists an amount SD (= 8Dg He SD) relative to the ground. Point G can lie anywhere in the plane. This means that a pencil of lines can be drawn through point F, and there are corresponding points of inter- sections E on the line $, in the range —% to + (see Fig. 5.7). If a point G is selected, such that the line FG is parallel to S,, then point E lies at infinity and the motion of the workpiece is a pure displacement alongThe stiffness mapping for a parallel manipulator 168 the ground for which 8D; = SDz = (4x, 6y; 0}. Point E must lie on the line $2. This is the condition of reciprocity for which the instant work must vanish: DE, Big = 8b5fa (VECa — XE Sa + Ta) = 0. (5.46) It follows from (5.45) that the law for the simultaneous control of motion and force can be expressed as 8Dg = GdDz + GrdDr, (5.47) where Gy and Gz are dimensionless gains for position and force errors. Sub- stitute (5.44) into (5.47) to yield 6Dg = Gr 8Dg — GK] diy = Gp6Dg — Gr8f[K]" Sa, (5.48) where i, = 6fSa. 5.5 Anote on the stability of spring systems A study of the instability of spring systems is beyond the scope of this text. However, it is of interest to note that some instability phenomena of the two-spring system were reported by Pigoski and Duffy (1993). Earlier relevant work on stability was reported by Haringx (1942, 1947, 1949, 1950), Eijk and Dijksman (1976), Dijksman (1979), and Eijk (1985). Stability phe- nomena that result in sudden erratic behavior can be explained by a mathe- matical theory called catastrophe theory (see Zeeman 1977, Arnol’d 1992, and Bruce and Giblin 1993). Catastrophe theory is currently being applied to the two-spring system by R. Hines (a research assistant at the University of Florida), D. Marsh (a lecturer in mathematics at Napier University, Edin- burgh), and the author. As far as the author is aware there is no study of the stability of the planar- three-spring system in progress. A relevant paper on the planar three-spring system was reported by Griffis and Duffy (1992). EXERCISE 5.1 The two-spring system (see Exr. Fig. 5.1) is in its unloaded configura- tion. A vertical force F is applied at C. Plot the locus of the equilibrium posi-Note on the stability of spring systems 169 x Br 2in. B, oO Zp = 2in k, = 2 Ubpiin or k, = 2 lbffin F bof Exercise Figure 5.1 tions for C when the magnitude of the force increases in increments of 0.25 Ibf in the range 0 = F =5 Ibf. (a) Assume that the spring matrix is given by ec c2l[kr O]fea Ko] = . i : | i | B | (b) Assume that the spring matrix is given by = 3, —82] [1 — pi) 0 —1 Cy {KI = (Ro) + [ C1 e2 ] [ 0 Al — ool [= aReferences Arnol’d, V. I. 1992. Catastrophe Theory. New York: Springer-Verlag. Ball, R. S. 1900. A Treatise on the Theory of Screws. Cambridge: Cambridge University Press. Bruce, J. W., and P. J. Giblin. 1993. Curves and Singularities. Cambridge: Cambridge University Press. Chung, Y. S., M. Griffis, and J. Duffy. 1994. “Repeatable Joint Displacement Generation for Redundant Robotic Systems.” Trans. ASME Journal of Mechanical Design 116, no. (1), (March):11-16. Dijksman, J. 1979. “A Study of Some Aspects of the Mechanical Behavior of Cross- Spring Pivots and Plate Spring Mechanisms with Negative Stiffness.” Ph.D. Disser- tation, Delft University of Technology, Amsterdam, The Netherlands. Dimentberg, F. M. 1968, The Screw Calculus and Its Applications in Mechanics. U.S. Department of Commerce FTD-HT-23- 1632-67. Eijk, J. 1985. “On the Design of Plate Spring Mechanisms.” Ph.D, Dissertation, Delft Uni- versity of Technology, Amsterdam, The Netherlands. Eijk, J., and J. Dijksman. 1976. “Plate Spring Mechanisms with Constant Negative Stiff- ness.” Vakgroep Fijnmechanische Techniek, Sune. Griffis, M., and J. Duffy. 1992. “Comparing Structures of Stiffness Matrices Using In- variants.” Presented at the VII CISM-IFToMM Symposium “Ro. Man. Sy *92.” Haringx, J. 1942. “On the Buckling and the Lateral Rigidity of Helical Compression Springs.” Proc. Ned. Akad. Wetenschappen, Amsterdam 45. Haringx, J. 1947. “On Highly Compressible Helical Spring and Rubber Bars and Their Application in Vibration Isolation.” Ph.D. Dissertation, Delft University of Tech- nology, Amsterdam, The Netherlands. Haringx, J. 1949, “Elastic Stability of Helical Springs at a Compression Larger than Orig- inal Length.” Applied Scientific Research A1, 417-434. Haringx. J. 1950, “Instability of Springs.” Phillips Technical Review, 11(8). 171References 172 Hunt, K. H. 1978 1990. Kinematic Geometry of Mechanisms. Oxford: Clarendon Press, 2-3, and 108-110. Klein, F. 1939, “Elementary Mathematics from an Advanced Standpoint. Geometry,” 25-66, and 86-88. New York: Macmillan. Li, S., and G, K. Matthew. 1987. “A Kinematic and Dynamic Investigation of the Planar Assure II Group.” In Proceedings of the Seventh World Congress on the Theory of Machines and Mechanisms, E. Bautista, J. Garia-Loman, and A. Navarro (Eds.). Ox- ford: Pergamon Press, 141-145. Pigoski, T., and J. Duffy. 1993. “An Inverse Force Analysis of a Planar Two-Spring Sys- tem.” Presented at the First Austrian IFToMM Symposium, Seggauberg, Austria (4-9 July). Published 1995. Trans. ASME Journal of Mechanical Design 117 (December). Reuleaux, F. 1876. The Kinematics of Machinery, p. 41. New York: Macmillan, Zeeman, E. 1977. Catastrophe Theory: Selected Papers, 1972-1977. Boston: Addison- Wesley Advanced Book Program.Index absolute displacement, 8, 9 coordinates axis, 89 force, 50 homogeneous, 43, 84 line, 40, 83, 85 Pliicker, 43 point, 40, 48 ray, 89 rotation, 56, 91 rotor, 82 translation, 56, 88, 90, 91 couple, 50 coupler, 8 crank, 9 degrees of freedom, 5, 11, 12 displacement analysis forward, 19, 25, 38 reverse, 19, 25, 36 end effector, 9 forces antiparallel, 51 components of, 50 173 intersecting, 55 magnitude of, 50 parallel, 54 4R mechanism, 8, 9, 107-110 frame, 7 Grassmann, 44, 48, 63, 64 in-parallel manipulator, 10, 11, 34, 36 kinematic analysis forward, 96, 143 reverse, 97, 143 kinematic chain, 6 closed, 7 open, 9 kinematic inversion, 7, 9, 10 lamina, 5 line differential of, 152 equation, 42 induced transformation of, 64 at infinity, 51, 53 pencil of, 54, 56 skew, 126Index 174 mechanism, 7 mobility, 11-13 one-system, 121 P kinematic pair, 6, 7 pole, 8 PPP manipulator, 23 reciprocal, 120 reciprocal one-system, 122, 124, 125 reciprocal two-system, 121, 123 redundant parallel manipulators, 80 redundant serial manipulators, 33, 106 relative displacement, 8, 9 R kinematic pair, 6, 7 rotor, 82 serial manipulator, 9 static analysis forward, 69, 131 reverse, 69, 131 triangle, area of, 44 trigonometrical equations, 15-18 two-system, 122 unbounded, 52, 53 vector free, 43, 51, 58, 83, 88, 91 line bound, 43, 50, 58, 91