See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/347119668

Determination of Kinematic Control Parameters of Omnidirectional AGV

Robot with Mecanum Wheels Track the Reference Trajectory and Velocity

Chapter · January 2021

DOI: 10.1007/978-3-030-64719-3_36

CITATIONS READS

6 148

4 authors, including:

Thai Nguyen Hong

42 PUBLICATIONS 113 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Roots blower View project

All content following this page was uploaded by Thai Nguyen Hong on 17 September 2022.

The user has requested enhancement of the downloaded file.

 DETERMINATION OF KINEMATIC CONTROL PARAMETERS
OF OMNIDIRECTIONAL AGV ROBOT WITH MECANUM
WHEELS TRACK THE REFERENCE
TRAJECTORY AND VELOCITY
Trinh Thi Khanh Ly1,*, Nguyen Hong Thai2, Le Quoc Dzung1, Nguyen Thi Thanh3
1,*
Faculty of Automation Technology, Electric Power University, Ha noi, Vietnam
2
Department of Mechanical Design and Robotics, School of Mechanical Engineering, Hanoi-
University of Science and Technology, Hanoi, Vietnam
3 Department of Control and Automation, Faculty Electrical Engineering, Economic and Tech-

nical University of Industry, Vietnam

Abstract.
Nowadays, in order to save the floor space of the production workshops, the
mecanum wheels have been applied to the design of Automated Guided Vehicles
(AGV) robot to create highly flexible and omnidirectional AGV robot. Omni-
directional robotic platforms have vast advantages over a conventional design in
terms of mobility in congested environments. These environments are commonly
found in factory workshops offices, warehouses, hospitals and elderly care facilities.
To be able to control these robots in any given trajectory, it is necessary to set up the
kinematic control parameters. The paper presents the method of establishing the kin-
ematic equation of omnidirectional robots by the centre of instantaneous velocity,
thereby determining the control parameters of AGV robot following a reference tra-
jectory. In addition, the authors also designed the robot's motion trajectories using the
NURBS curve to solve the robot's trajectory planning problem according to different
application scenarios of modern industrial production.

Keywords: AGV robot, Kinematic of AGV, Omnidirectional mecanum wheel,

Trajectory design, NURBS curve.

1 Introduction

Mecanum omnidirectional wheels described in Fig. 1 were invented by a Swedish

engineer Ilon in 1973 at Mecanum Company [1] and applied to the design of the first
AGV in 1997. When torque is transmitted to the wheels, the rollers on the wheel that
come into contact with the floor form two velocity components: in the direction of the
wheel's movement and is perpendicular to the roller axis and depends on the direc-
tion of applied torque. Therefore, when controlling the whole motor, the wheels will
form a resultant pushing AGV robot in different directions, increasing the flexibility
of the robot.
Therefore, in recent years, this wheel has also been used in the design of AGV ro-
bots for logistics, transportation in tight spaces, which is not enough to design the
turning path for AGV. such as the ominidirection AGV model of L. Schulze at al [2]
2

using mecanum wheels with the conveying and towing functions described in Fig. 2,
or that of Michael Göller at al [3, 4] with mecanum wheels for supermarket custom-
ers. In addition, there are a number of other studies [5, 6] that show different AGV
robots for different industrial production application scenarios.

L
M 

 
VL V
Fig. 1. Mecanum wheels Fig 2. Omnidirection AGV robots using Mecanum
wheels [2]
To kinematically control the omnidirection AGV robot in the desired orbit, the first
thing is to set up the kinetic model of the AGV robot. In terms of this, Gfrerrer [7]
and Tatsuro [8] determined the effect of roller size, wheels and number of rollers on
the speed of robots, or another research of Hamid [9] and Lin [10] using the matrix
method and the non-slip condition of the wheel on the floor to establish forward and
inverse kinematic model, thereby designing adaptive controller.
That has led to too many parameters in the kinematic equation and such a complex
problem, to overcome the above disadvantage, in this article the authors set up the
inverse kinematic equations directly with instantaneous center of velocities. On that
basis, the parameters of dynamically controlling ominidirection AGV robots by the
reference trajectories and still achieve the desired speed of the robot. In addition, the
article also presents the direction design method for robots with Non-uniform rational
B-spline (NURBS) in the general case.

2 Determination of kinematic control parameters of

omnidirection AGV robot with mecanum wheels

2.1 Establishment of Kinematic equations of omnidirectional AGV

robot with Mecanum wheels

In order to establish the kinematic equation of AGV robot, we let J f {O f x f y f } as a

fixed reference frame attached to the floor of the robot; J R {GR xR yR } is a reference
system mounted on AGV robot (see Fig. 3);
 3

 
V2 V
L2
y 2
f y  xR d
VG 

R  V
 VG x 3
 V1 VG y 3
VL1 
VL 3
y

G
1  GR V '3
 V
V4 L 4
L
4 r
{}

xG xf
Of
Fig. 3. Diagram of AGV robots with Mecanum wheels

û is the robot positioning parameters in the frame  f ,

T
q f = éêëx f (t ) y f (t ) f (t )ù
ú
T
and q R = [xR (t ) yR (t ) f (t )] is the parameter in the frame R , if
T
q f = éêëx f (t ) y f (t ) f (t )ù
û is in the frame of J f {O f , x f , y f } , we can determine the
ú
control parameter q&R of Robot with:
q&R = Q(f )T q&f (1)
Where:
écos f - sin f 0ù
ê ú
Q(f ) = êêsin f cos f 0ú
úis a directing cosine matrix of the reference frame
ê 0 0 1ú
ë û
J R to the reference frame J f .

For the reference frame J R of robot: To the wheel No. 1 (Fig. 3), we can see:

ìï
ïï V1 (t ) + VL1 cos g = VGy (t ) - L W(t )
ïï R
2
í (2)
ïï d
ïï VL1 sin g = VGxR (t ) - W(t )
ïî 2
4

Where: L, d is the distance between the two wheels, the distance between the front
and rear wheels; Ω (t) is the angular velocity of AGV robot; V1 (t ) is the speed of
wheel 1 such as V1 (t ) = r w1 (t ) where w1 (t ) is the angular velocity of wheel No.1 and
r is the radius of the wheels (see Fig. 4)


r

i
i

Vi

 
VLi
Fig. 4. The velocity components on the Mecanum wheels

De-velocity of roller VL1 in equation (2), we get the velocity of wheel No.1 as fol-
lows:
L æ d ö 1
V1 (t ) = VGy R (t ) - W(t ) + ççVGxR (t ) - W(t )÷÷÷ (3)
2 ç
è 2 ø tgg
We apply the same method to wheels No.2, 3, 4, after all we have:

ìï
ïï V (t ) = V (t ) - L W(t ) + æ ö
ççVGx (t ) - d W(t )÷÷ 1
ïï 1 GyR
2 ç
è R
2 ÷
ø tg g
ïï
ïï L æ d ö 1
ïï V2 (t ) = VGyR (t ) + W(t ) - çççVGxR (t ) - W(t )÷÷÷
ïï 2 è 2 ø tg g
í
ïï L æ d ö 1
ïï V3 (t ) = VGyR (t ) + W(t ) + ççVGxR (t ) + W(t )÷÷÷
ïï 2 çè 2 ø tg g
ïï
ïï L æ d ö÷ 1
ç
ïï V4 (t ) = VGyR (t ) - 2 W(t ) - ççèVGxR (t ) + 2 W(t )ø÷÷tg g
ïî (4)
o
g = 45
To change the direction of the omnidirection AGV in any direction. Set in
(4) and change to the algebraic form, we have:
é1 1 1 ù
ê - ( L + d )ú
êr r 2r ú
éw1 (t ) ù ê ú
ê ú ê 1 1 1 úé ù
êw2 (t )ú ê- ( L + d ) úêVGxR (t )ú
ê ú ê r r 2r úê ú
êw (t ) ú= ê úêVGyR (t )ú
ê1 1 1 úê
ê3 ú
êw (t )ú ê ( L + d ) úê W(t ) úú
êë 4 úû êr r 2r úë û
ê 1 1 1 ú
ê- - ( L + d )úú
ê
ë r r 2r û (5)
 5

é1 1 1 ù
ê - ( L + d )ú
êr r 2r ú
ê ú
ê 1 1 1 ú
ê- (L + d ) ú
ê ú
Let: ω = [w1 (t ) w2 (t ) w3 (t ) w4 (t )] and J = ê r r 2r
T
ú
ê1 1 1 ú
ê (L + d ) ú
êr r 2r ú
ê 1 1 1 ú
ê- - ( L + d )ú
ê ú
ë r r 2r û
Eq. 5 is rewritten as follows:
ω = Jq&R (6)

2.2 Determination of kinematic control parameters by the given

trajectories and velocity

 ( x f (t ), y f (t )) and VG
T
With orbit , we determine q&f = éêëx&f (t ) y&f (t ) f&(t )ù
û in
ú
the reference frame J f , then substitute into eq. 1 and combine with eq. 6 we have:
ω = JQT (f )q&f (7)
From (7) we determine the control parameters of the matching variables on the
mecanum wheels:
ìï
ïï w (t ) = 1 æ ö
çça(t ) x& (t ) + b(t ) y& (t ) - 1 ( L + d )f&(t )÷÷
ïï 1 rè ç f f 2 ÷
÷
ø
ï ïï
ïï 1 æç & 1 ö
ïï w (t ) = çça(t ) y (t ) - b(t ) x& (t ) + ( L + d )f&(t )÷÷÷÷ (8)
ïï 2 rè f f 2 ø
í
ïï 1 æçæç & ö 1 ö
ïï w (t ) = ççx (t )a(t ) + b(t ) y& (t )÷÷÷+ ( L + d )f&(t )÷÷÷
ïï 3 r çèçè f f ø 2 ø÷
ïï
ïï
1 æç 1 ö
ïï w (t ) =
ïï 4 çç- b(t ) x& (t ) + a(t ) y& (t ) - ( L + d )f&(t )÷÷÷÷
ïî rè f f 2 ø
ìï a(t ) = cos f (t ) - sin f (t )
Where: ïí
ïïî b(t ) = cos f (t ) + sin f (t )
Eq. 8 determines the control parameters of the angular velocity of the driven mo-
tors on the Mecanum wheel axle so that the omnidirection AGV robot follows the
trajectory and achieves the given velocity.

3 Design of moving trajectory of AGV robots

The trajectory of AGV robots is usually designed as straight lines along the corri-
dors used as the paths of the production line [11, 12] and at the intersections softened
by curves by robot’s breakdown velocity and safety corridor. This will modelize the
trajectory into a fixed map in the memory of the robot. To make it easier to design
6

and encode the robot's route map, we use the coordinate matrix as a database, and the
interpolation will be done by the NURBS to guide the robot well. According to the
work [13] the non-uniform B - Spline rational curve (NURBS) is defined by:
n+ 1
P(t ) = å Bi Ri , k (t ) (9)
i= 1

Where B i is the vertices of the interpolated polygons in 2D space, and Ri , k (t ) is

the basis function of the B-Spline rational curve and is given by:
h N i , k (t )
i
Ri , k (t ) = n+ 1
(10)
å hi N i , k (t )
i= 1

As: hi  0 for all values of i, N i , k (t ) given by the recursive formula:

ìï
ïï N (t ) = ìïïí 1 xi £ t £ xi + 1
ïï i ,1 ïïî 0
ïí (11)
ïï (t - xi )Ni,k - 1 (t ) (xi+ k - t )Ni+ 1,k - 1 (t )
ïï Ni ,k (t ) = +
ïï xi + k - 1 - xi xi + k - xi + 1
î
xi is the value of the node vector and satisfies the condition xi  xi 1 . The tan-
gent to the curve is given by:
é n+ 1 ù
ê ú
ê & hi N i , k (t )å hi N&i , k (t ) ú
n+ 1
ê hi Ni , k (t ) ú (12)
P&(t ) = å Bi ên+ 1 - i= 1
ú
ê æ n+ 1 ö
2
ú
êå hi Ni , k (t ) çç h N (t )÷
i= 1

çèåi= 1 i i , k ÷ ÷ úú
÷
ê i= 1 ø û
ë
Thus, at each coordinate point K(xf, yf) on the interpolation curve {} can we com-

pletely determine the tangent vector  .

4 Simulation examples

Applied to omnidirectional AGV robots with dimension parameters: radius of

mecanum r = 5 cm; The distance between the two wheels L = 50 cm; Distance be-
tween front and rear wheel d = 80 cm; With the assumption: (1) Considering at the
time the omnidirection AGV robot moves at a steady velocity VG = 0.3 m/s; (2) there
is no horizontal and vertical sliding of robots on the road surface. The coordinates of
the grid node on the trajectory are given in Table 1 below.
 7

Table 1. Coordinates of polygon nodes designing trajectory of the vehicle.

B1 B2 B3 B4 B5 B6 B7

0   6.5  6.5  3  3  6.5  6.5

0   0   5  5 8  8   13 
             
B8 B9 B10 B11 B12 B13

 6 .5   6 .5  3  3  6 .5   6 .5 
 13  8 8 5 5 0
           
Fig. 5 is the roadmap of the AGV robot and Fig. 6 is the trajectory of the robot af-
ter interpolation with NURBS curve.
yf y [m]
B7 B8
13
3
12

10

B5 B10
B6 B9 8

2 4
6 { }
B3 B4 B12
B11 4

B1 xf 1 x
B2 Of B13 -80 -6 -4 -2 0 2 4 8
[m]
6
Fig. 5. Roadmap of nodes Fig. 6. Robot's trajectory is interpolated
on the trajectory of AGV robot with NURBS curves
Corresponding to each point K(xf, yf) on the trajectory {ξ} in the reference frame

J f we determine the tangent vector  through eq. 12. Attach the center coordinate
GR(xG,yG) of the AGV robot to the point K(xf, yf ) and the yR axis coincide with

 then the angle f (t ) is determined:
r
f (t ) = Ð(t (t ), y f ) (13)
And f&(t ) is found by
V
f&(t ) = W(t ) = G (14)
r (t )
Where ρ(t) is the curvature radius of {ξ} corresponding to the point K(xf, yf) at time
t and depicted in Fig. 7, and the graph depicting the angular velocity Ω(t) of the robot
(The speed changes by the direction in the trajectory) is shown in Fig. 8.
8

35 0.3
1 3 1
30 0.2
 (t) (m)

25

 (rad/s)
0.1

1 2 3 4 1
20 0

15
-0.1
10
2 4 -0.2
5 t(s)
0
0 0 -0.3 50 100 150 200
50 100 150 200
t (s)
Fig. 7 Turning center radius on the Fig.8. Velocity changing robot
moving path of AGV Robot change direction

2 3 4
10
1 [rad/s]

t (s)
0 20 40 60 80 100 120 140 160 180 200

10
2 [rad/s]

t (s)
0 20 40 60 80 100 120 140 160 180 200
0
10
3 [rad/s]

t (s)
0 20 40 60 80 100 120 140 160 180 200
0
10
4 [rad/s]

t (s)
0 20 40 60 80 100 120 140 160 180 200
0
Fig. 9. The angular velocity of the wheels as the AGV Robot moves along the trajectory
Insert eqs 13, 14 and 15 into eq.8 to determine the angular velocity ꞷ1(t), ꞷ2(t),
ꞷ4(t), ꞷ4(t), so that the AGV robot can follow the given trajectory and achieve the
velocity VG = 0.3 m/s. Fig. 9 is a graph of angular velocity of four wheels. On the
other hand, from the previous VG velocity we have:
 9

ìï x&f (t ) = VG sin f (t )
ïí (15)
ïï y&f (t ) = VG cos f (t )
î
From Figs. 5 to 9, we find that:
i) When making the trajectory map of AGV robot with NURBS interpolation
method through grid nodes. As a result, the database that stores the robot's path in
memory is greatly reduced. Instead of having to remember all the data for the path,
the route designer just needs to enter the grid nodes.
ii) When the AGV Robot travels from its initial position "1" to "2", the robot's rota-
tion in the trajectory changes from clockwise to counter-clockwise (see Fig. 8) and
this cycle repeats 4 times according to the symmetry of the moving trajectory. Espe-
cially when passing through position "2" and "3", it suddenly changes direction, the
angular velocity of the robot has a greater variation at the position of "2" and "4"and
much greater than the positions "1" and "3". The reason is that the instantaneous radi-
al radius at positions "2" and "4" is the smallest (see Fig. 7). Therefore, in order not to
have a sudden change of direction when designing the trajectory, it is necessary to
avoid small turning radius and at these points the wheels get sliding phenomenon on
the road surface causing errors of position and direction.

4. CONCLUSION

This study has established four inverse kinematic control parameters for omnidi-
rection AGV robots with mecanum wheel to ensure that the robot follows the given
trajectory and achieves the desired velocity. Using the grid of coordinates as a data-
base to store the robot's road map will reduce the memory capacity of the robot and
the interpolation with the NURBS curve will be softer and more flexible than the
studies previously finding equivalent arcs through the problem of breakdown velocity
and the safety corridor of the workshop. In addition, the study also showed that at the
locations where the robot changes direction, it is necessary to determine the minimum
radius of curvature so that sliding does not occur on the road surface, causing the
position errors of robots. This is an issue that authors will continue to publish in the
near future.

ACKNOWLEGEMENT
This research was funded by the Ministry of Industry and Trade in a ministerial-
level scientific and technological research project, conducted in 2020, code:
DTKHCN.076/20

References

1. Ioan Doroftei, Victor Grosu, Veaceslav Spinu.: Omni directional Mobile Robot – Design
and Implementation; Bioinspiration and Robotics Walking and Climbing Robots (2017)
511- 528.
 10

2. L. Schulze, Sebastian Behling, Stefan Buhrs; Development of a Micro Drive - Under

Tractor Research and Application; Proceedings of the International MultiConference of
Engineers and Computer Scientists, Vol. II, March 16-18, 2011, Hong-Kong.
3. Michael Göller, Thilo Kerscher, Johann Marius Zöllner, Rüdiger Dillmann, Michel Devy,
Thierry Germa, Frédéric Lerasle; Setup and Control Architecture for an Interactive
Shopping Cart in Human All Day Environments; In Proceeding of the Int. Conf. on
Advanced Robotics (2009) 1-6.
4. Michael Göller, Thilo Kerscher, Marco Ziegenmeyer, Arne Ronnau, Johann Marius
Zöllner , Rüdiger Dillmann; Haptic Control for the Interactive Behavior Operated
Shopping Trolley InBOT; International Conference on Advanced Robotics (2009).
5. Diogo Puppim de Oliveira, Wallace Pereira Neves dos Reis, Orides Morandin Junior; A
Qualitative Analysis of a USB Camera for AGV Control; Sensors (2019) Doi:
10.3390/s19194111.
6. A. Gfrerrer; Geometry and kinematics of the Mecanum wheel; Computer Aided Geometric
Design 25 (2008) 784–791.
7. Tianran Peng, Jun Qian, Bin Zi, Jiakui Liu, Xingwei Wang; Mechanical Design and
Control System of an Omni-directional Mobile Robot for Material Conveying; Procedia
CIRP 56, 412 – 415 (2016).
8. Tatsuro Terakawa, Masaharu Komori, Yuuta Yamaguchi, Yugo Nishida.: Active omni
wheel possessing seamless periphery and omnidirectional vehicle using it; Precision
Engineering 56, 466–475 (2019).
9. Hamid Taheri, Bing Qiao, Nurallah Ghaeminezhad.: Kinematic Model of a Four Mecanum
Wheeled Mobile Robot; International Journal of Computer Applications Volume 113 –
No. 3, 0975 – 8887 (2015).
10. Lih-Chang Lin, Hao-Yin Shih.: Modeling and Adaptive Control of an Omni-Mecanum-
Wheeled Robot; Intelligent Control and Automation 4, 166-179 (2013).
11. Jung Hoon Lee, Beom Hee Lee, Myoung Hwan Choi.: A Real-Time Traffic Control
Scheme of Multiple AGV Systems for Collision Free Minimum Time Motion: A Routing
Table Approach; IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems
and Humans, Vol 28, No 3, 347 – 358 (1998) .
12. Gerald Cook.: Mobile Robots Navigation Control and Remote Sensing. Wiley (2011).
13. David F. Rogers.: An Introductron to NURBS. Morgan kaufmann Publishers (2001).

View publication stats