ROBOTICS : BASICS AND A DVANCED C ONCEPTS

V ELOCITY AND S TATIC A NALYSIS OF M ANIPULATORS

Ashitava Ghosal

Department of Mechanical Engineering

&
Centre for Product Design and Manufacturing
Indian Institute of Science
Bangalore 560 012, India
Email: [email protected]

NPTEL, 2020

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 1 / 103
 1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 2 / 103
O UTLINE
1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 3 / 103
I NTRODUCTION

Position kinematics → Position & orientation of links, workspace, mobility etc.

Change of position and orientation with respect to time → Velocity kinematics
Linear velocity as derivative of position vector.
Angular velocity in terms of derivative of rotation matrix.
Topics in velocity kinematics include
Linear and angular velocities of links
Manipulator Jacobian(s)
Singularities in velocity domain
Static equilibrium
Relation between external forces & moments and joint torques & forces.
Singularities in force domain

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 4 / 103
O UTLINE
1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 5 / 103
L INEAR V ELOCITY OF R IGID B ODY
The linear velocity of Oi with respect to {0} is defined as
d 0 0 O (t + ∆t) −0 O (t)
0 ∆ i i
VOi = Oi (t) = lim (1)
dt
Rigid body at
∆t→0 ∆t
t + ∆t
Ẑ
{0} ‘0’ denote the coordinate system {0}
{i}(t)
{i}(t + ∆t)
where the limit is taken.
0
Oi (t)
Oi The linear velocity vector can be
described in {j} as
Oi ( )
Ŷ
j 0
VOi = j0 [R]0 VOi (2)

0
Oi (t + ∆t) Two different coordinate system
X̂
involved: where differentiation done,
Rigid body at t
and where described!
F IGURE : Linear velocity of a rigid body . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 6 / 103
A NGULAR V ELOCITY OF R IGID
B ODY
Angular velocity cannot be obtained as a time derivative of 3 quantities (say 3
Euler angles) representing orientation.
Angular velocity from time derivative of rotation matrix.
Recall
T
0
i [R] 0i [R] = [U], [U] is a 3 × 3 identity matrix
Differentiate with respect to time t
˙ 0 [R]T
0 [R] ˙ T = [0]
+ 0i [R] 0i [R]
i i
Note: derivative of a matrix implies derivative of all components of the matrix.
Above equation can be written as
˙ 0 [R]T
0 [R] ˙ 0 [R]T )T = [0]
+ (0i [R]
i i i

Define a 3 × 3 skew symmetric matrix

0 ∆0 ˙ 0 T
i [Ω]R = i [R] i [R] . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 7 / 103
A NGULAR V ELOCITY OF R IGID
B ODY – S KEW S YMMETRIC M ATRIX
Skew-symmetric matrix in detail
 
0 −ωzs ωys
0
i [Ω]R =
 ωzs 0 −ωxs  (3)
−ωys ωxs 0
Product of O [Ω]
i R and (px , py , pz )T ∈ ℜ3 is a cross-product
 s 
ωy pz − ωzs py
0 T
i [Ω]R (px , py , pz ) =
 ωzs px − ωxs pz  = 0 ω si ×0 p (4)
ωxs py − ωys px
0 [Ω]
i Rcalled angular velocity matrix – 0 ω si : angular velocity vector of {i} with
respect to {0}.
In contrast to linear velocity, angular velocity vector is not a straightforward
differentiation of orientation variables! . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 8 / 103
A NGULAR V ELOCITY IN TERMS OF
E ULER ANGLES
Angular velocity in terms of Z-Y-Z Euler angles.
Recall for α , β and γ as the Z-Y-Z Euler angles
 
cα cβ cγ − sα sγ −cα cβ sγ − sα cγ cα sβ
A
B [R] =  sα cβ cγ + cα sγ −sα cβ sγ + cα cγ sα sβ  (5)
−sβ cγ sβ sγ cβ

Obtain A ˙ A T
B [R] B [R]
The X , Y and Z components of the angular velocity vector

ωxs = γ̇ cos α sin β − β̇ sin α

ωys = γ̇ sin α sin β + β̇ cos α (6)
ωzs = γ̇ cos β + α̇
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 9 / 103
A NGULAR V ELOCITY – S PACE AND
B ODY F IXED

0 [Ω] T
i R — Derived from right multiplication 0i [R] 0i [R] = [U].
Derived 0 ω si called the space-fixed angular velocity — Superscript s.
0 [R]T 0 [R] = [U] → Another skew-symmetric matrix
i i
 
0 −ωzb ωyb
0 ∆0 T ˙ =  ωb
0 [R] 0 −ωxb  (7)
i [Ω]L = i [R] i z
−ωyb ωxb 0

0 [Ω] T 0
i L derived from left multiplication 0i [R] i [R] = [U]
Angular velocity vector 0ω b from the three components (ωxb , ωyb , ωzb ).
i

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 10 / 103
A NGULAR V ELOCITY – C ONTD .
For the Z-Y-Z rotation the three components are

ωxb = −α̇ cos γ sin β + β̇ sin γ

ωyb = α̇ sin β sin γ + β̇ cos γ (8)
ωzb = α̇ cos β + γ̇
0ω bcalled body-fixed angular velocity vector of {i} with respect to {0} —
i
Superscript b.
The two skew-symmetric matrices are related like two tensors
0 T
i [Ω]R = 0i [R] 0i [Ω]L 0i [R] (9)

The two angular velocities are related as

0
ω si = 0i [R]0 ω bi (10)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 11 / 103
A NGULAR V ELOCITY OF R IGID
B ODY (C ONTD .)
{0} Ẑ Rigid Body
at t

Consider rigid body undergoing pure

i
p
rotation about a fixed point.
Points 0 Oi (t) and 0 Oi (t + ∆t) are
Ŷ coincident and only the elements of
{i}t the rotation matrix 0i [R] change with
Oi time.
X̂ Point P located by i p, and fixed in
Rigid Body at
t + ∆t {i}.
{i}t+∆t

F IGURE : Angular velocity of a rigid body . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 12 / 103
A NGULAR V ELOCITY OF R IGID
B ODY (C ONTD .)
Location of P in {0} 0 p = 0i [R]i p
Since P is fixed in {i}
0 ∆ ˙ ip
ṗ = 0 Vp = 0i [R]
−1 T
and since 0i [R] = 0i [R] ,

0
Vp = ˙ 0 [R]T 0 p
0 [R]
i i

i [Ω]R p = ω i ×0 p
0 0 0 s
= (11)

The coordinate system {i} does not appear except in denoting that rigid body
{i} is being considered.
Space-fixed angular velocity vector is said to be independent of the choice of
the body coordinate system. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 13 / 103
A NGULAR V ELOCITY OF R IGID
B ODY (C ONTD .)
Using relation between 0i [Ω]R and 0i [Ω]L
0 0 0 0 T 0
Vp = i [R] i [Ω]L i [R] p = 0i [R]0i [Ω]L i p
−1 0
to get 0i [R] V p = 0i [Ω]L i p
Yielding
i
Vp = 0i [Ω]L i p = 0 ω bi ×i p (12)
Except for denoting the reference (fixed) coordinate system, the coordinate
system {0} does not appear!
Body-fixed angular velocity vector is said to be independent of the choice of the
fixed coordinate system.
Unless explicitly stated, space-fixed angular velocity vector derived from
˙ 0 [R]T is normally used in kinematic analysis.
0 [R]
i i . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 14 / 103
A NGULAR V ELOCITY IN S ERIAL
M ANIPULATOR – ROTARY (R) JOINT
For two links connected by a rotary (R) joint

i [R(k̂, θi )]
0
i [R] = 0i−1 [R] i−1

The time derivative operation

˙ 0 [R]T d 0
i [R(k̂, θi )]) i [R(k̂, θi )]
i−1 0
0 [R]
i = ( [R] (i−1 T T
i−1 [R] )
i
dt i−1
Rewrite above equation as

i [Ṙ(k̂, θi )] i [R(k̂, θi )] ) i−1 [R]

0
i [Ω]R = 0i−1 [Ω]R + 0i−1 [R] (i−1 i−1 T 0 T

(i [K ]θi ) where i−1 [K ] is the

i−1
i [R(k̂, θi )] = e
Simplify using the result i−1 i
skew-symmetric form of the rotation axis vector k̂ and θi is the rotation at the
R joint . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 15 / 103
A NGULAR V ELOCITY P ROPAGATION
IN S ERIAL M ANIPULATOR – R JOINT
d (i [K ]θi ) i−1 i−1
[K ]θi )
k̂ is fixed in {i − 1} and {i} → dt i [K ]θ̇i e
= i−1
e (i

From above and properties of a rotation matrix,

0
i [Ω]R = 0
i−1 [Ω]R + 0i−1 [R] i−1
i [K ] 0
i−1 [R]
T
θ̇i
= 0
i−1 [Ω]R + 0i [K ]θ̇i
and in terms of the space-fixed angular velocity 0 ω (·)
0
ω i = 0 ω i−1 +0 k̂i θ̇i , R joint along Z − axis
Pre-multiply both sides by i0 [R] and simplify to get
i
ω i = ii−1 [R]i−1 ω i−1 + θ̇i (0 0 1)T (13)
iω denotes i0 [R]0 ω i – i ω i not necessarily 0.
i
Equation (13) gives the angular velocity propagation in links of a serial
manipulators connected by R joints. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 16 / 103
L INEAR V ELOCITY P ROPAGATION IN
S ERIAL M ANIPULATOR – R JOINT
For two consecutive links in a serial manipulator
0
Oi =0 Oi−1 + 0i−1 [R]i−1 Oi

Taking derivatives with respect to time on both sides

0
VOi =0 VOi−1 +0 ω i−1 × 0i−1 [R]i−1 Oi

Simplify and rewrite above as

i
Vi = ii−1 [R](i−1 Vi−1 +i−1 ω i−1 ×i−1 Oi ) (14)

Note: i Vi and i−1 Vi−1 denote i0 [R]0 Vi and i−1 0

0 [R] Vi−1 , respectively. They are
not necessarily 0!
Equation (14) gives the linear velocity vector propagation in links of a serial
manipulator connected by rotary joints. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 17 / 103
V ELOCITY P ROPAGATION –
P RISMATIC (P) J OINTS
Two links connected by a prismatic (P) joint
Prismatic joint allows relative translation between {1 − i} and {i} → Angular
velocity is same.
Relative translation is along Z − axis → d˙i (0 0 1)T
Velocity propagation for P joint
Angular velocity
i
ω i = ii−1 [R]i−1 ω i−1 (15)
Linear velocity
i
Vi = ii−1 [R](i−1 Vi−1 +i−1 ω i−1 ×i−1 Oi ) + d˙i (0 0 1)T
(16)
i [R]i−1 ω =∆ i ∆ i
where i−1 i ωi and i [R]i−1 V =
i−1 i Vi . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 18 / 103
 V ELOCITY P ROPAGATION – P LANAR
3R MANIPULATOR
X̂3 , X̂T ool

{T ool} X̂2
Link 3 θ3
ŶT ool l3 All joint axis are parallel and coming
O3
outwards.
{3} {0} is fixed →
Ŷ3 Link 2
Ŷ0

Ŷ1 Ŷ2 {2}

l2
θ2 X̂1
0
ω0 = 0
0
{0} V0 = 0
{1} O2
l1 Link 1 Links connected by rotary (R) joint
θ1 → Equations (13) and (14) give
X̂0
velocities of all links.
O1

F IGURE : The planar 3R manipulator . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 19 / 103
V ELOCITY P ROPAGATION – P LANAR
3R MANIPULATOR
For i=1
1
ω 1 = (0 0 θ̇1 )T , 1
V1 = 0
For i=2
2
ω 2 = (0 0 θ̇1 + θ̇2 )T
    
c2 s2 0 0 l1 s2 θ̇1
2
V2 =  −s2 c2 0   l1 θ̇1  =  l1 c2 θ̇1 
0 0 1 0 0
For i=3
3
ω 3 = (0 0 θ̇1 + θ̇2 + θ̇3 )T
 
(l1 s23 + l2 s3 )θ̇1 + l2 s3 θ̇2
3
V3 =  (l1 c23 + l2 c3 )θ̇1 + l2 c3 θ̇2 
0 . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 20 / 103
V ELOCITY P ROPAGATION – P LANAR
3R MANIPULATOR
For i = Tool
Tool
ω Tool = (0 0 θ̇1 + θ̇2 + θ̇3 )T
 
(l1 s23 + l2 s3 )θ̇1 + l2 s3 θ̇2
Tool
VTool =  (l1 c23 + l2 c3 + l3 )θ̇1 + (l2 c3 + l3 )θ̇2 + l3 θ̇3 
0
Linear and angular velocity in {0}
0
ω Tool = (0 0 θ̇1 + θ̇2 + θ̇3 )T (17)
and
 
−l1 s1 θ̇1 − l2 s12 (θ̇1 + θ̇2 ) − l3 s123 (θ̇1 + θ̇2 + θ̇3 )
0
VTool =  l1 c1 θ̇1 + l2 c12 (θ̇1 + θ̇2 ) + l3 c123 (θ̇1 + θ̇2 + θ̇3 )  (18)
0
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 21 / 103
S UMMARY

Linear velocity of a point on rigid body – Time derivative of position vector.

Angular velocity of a rigid body in terms of derivative of rotation matrix – Two
kinds
0 ∆ ˙ 0 T 0 T
i [Ω]R = 0i [R] i [R] — Obtained from time derivative of [R] 0i [R] = [U].
s i
i [Ω]R gives rise space fixed angular velocity vector ω i .
0 0

0 ∆ T ˙ — Obtained from time derivative of 0 [R]T 0 [R] = [U].

i [Ω] = 0 [R] 0 [R]
L i i i i
i [Ω]L gives rise body fixed angular velocity vector ω i .
0 0 b

Space fixed angular velocity vector used typically.

Propagation of linear and angular velocities between links.
Can easily obtain linear and angular velocity of any serial manipulator
connected with rotary (R) and prismatic (P) joints.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 22 / 103
O UTLINE
1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 23 / 103
S ERIAL M ANIPULATOR JACOBIAN
M ATRIX
Linear and angular velocity of {Tool}
0
ω Tool = (0 0 θ̇1 + θ̇2 + θ̇3 )T (19)

and  
−l1 s1 θ̇1 − l2 s12 (θ̇1 + θ̇2 ) − l3 s123 (θ̇1 + θ̇2 + θ̇3 )
0
VTool =  l1 c1 θ̇1 + l2 c12 (θ̇1 + θ̇2 ) + l3 c123 (θ̇1 + θ̇2 + θ̇3 )  (20)
0
In a compact form as
 
−l1 s1 − l2 s12 − l3 s123 −l2 s12 − l3 s123 −l3 s123
 l1 c1 + l2 c12 + l3 c123 l2 c12 + l3 c123 l3 c123 
  
 0 0 0  θ̇1
 
0
VTool =
 −− −− −−   θ̇2 
 (21)
 0 0 0  θ̇3
 
 0 0 0 
1 1 1
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 24 / 103
S ERIAL M ANIPULATOR JACOBIAN
M ATRIX
 0V

Tool
is a 6 × 1 entity – 0 VTool =  
∆
0V
Tool −−
0ω
Tool
0V is not a 6 × 1 vector1 – different units for velocities!
Tool
Use ‘–’ or ‘;’ to separate the linear and angular velocities & to remind that
Tool or ( VTool ; ω Tool ) is not a vector.
0V 0 0 T

Matrix in square brackets, 0Tool [J(Θ)], is called the Jacobian matrix for the
planar 3R manipulator.
0
Tool [J(Θ)] relate the linear and angular velocities of the {Tool} with the joint
velocities.
Jacobian matrix is for the end-effector or the {Tool} — See subscript Tool.
Linear and angular velocities are in {0} — See leading superscript 0.

. . . . . . . . . . . . . . . . . . . .
1 In theoretical kinematics, (0 ω Tool ;0 VTool ) is called twist. . . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 25 / 103
P ROPERTIES OF M ANIPULATOR
JACOBIAN MATRIX

0 is not a “true Jacobian” matrix

Tool [J(Θ)]
0
Tool [J(Θ)] is not obtained by direct differentiation of a vector-valued function,
The first and the last three rows represent linear and angular velocity,
Elements of the first three rows have units of length, elements of last three rows
have no units.
Similar to 0 VTool , top and bottom halves of 0
Tool [J(Θ)] matrix are separated by
‘–’.
Many matrix operations on 0Tool [J(Θ)] makes no sense – Finding the condition
number2 of this matrix is meaningless since it changes with the choice of length
units.

2 The condition number of a matrix is the ratio of the absolute value of the largest to the
. . . . . . . . . . . . . . . . . . . .
smallest eigenvalues. . . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 26 / 103
P ROPERTIES OF M ANIPULATOR
JACOBIAN MATRIX (C ONTD .)

0
Tool [J(Θ)] is a map 0
Tool [J(Θ)] : Θ̇ →0 VTool .
The manipulator Jacobian matrix can be derived3 for any serial manipulator
with rotary and prismatic joints
Compute the linear and angular velocities using propagation equations.
Rearrange in a matrix equation as done for the planar 3R manipulator.
0 is very important in velocity kinematics of serial manipulators.
Tool [J(Θ)]

3 TheJacobian matrix is defined for any differentiable vector-valued function X = Ψ(θ1 , . . . , θn ).

The Jacobian matrix, [J(Θ)], is the matrix of first partial derivatives of Ψ with respect to θi —
The i th column of [J(Θ)] is ∂∂Ψ
. . . . . . . . . . . . . . . . . . . .
θ . i . . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 27 / 103
P ROPERTIES OF M ANIPULATOR
JACOBIAN MATRIX (C ONTD .)
The elements of the Jacobian matrix are non-linear functions of the joint
variables Θ.
Manipulator in motion → 0Tool [J(Θ)] is time varying.
At instant with Θ known, 0Tool [J(Θ)] relates linear and angular velocities to joint
rates.
The relationship between joint rates and linear/angular velocities is linear!
The Jacobian matrix can be obtained for any link — Most often obtained for
end-effector.
The Jacobian matrix is always with respect to a coordinate system — Where
the linear and angular velocities are obtained.
Most often Jacobian matrix is with respect to fixed {0}.
Jacobian matrix can be written in any coordinate system using rotation
matrices. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 28 / 103
P ROPERTIES OF M ANIPULATOR
JACOBIAN MATRIX (C ONTD .)

The Jacobian matrix is m × n, where m is dimension of the motion space and n

is the number of actuated joints.
If 0Tool [J(Θ)] is square, i.e., m = n, and if the determinant det(0Tool [J(Θ)]) ̸= 0,
then
−1
Θ̇ = 0Tool [J(Θ)] 0 VTool (22)
Above relationship gives joint velocities required for a desired linear and angular
velocities of {Tool}.
Direct velocity kinematics — 0 VTool =0Tool [J(Θ)]Θ̇
−1 0
Inverse velocity kinematics — Θ̇ = 0Tool [J(Θ)] V Tool

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 29 / 103
 G EOMETRIC I NTERPRETATION OF
MANIPULATOR JACOBIAN MATRIX
X̂2
Consider a planar 2R manipulator shown in
Figure.
The linear velocity V of the end-effector
(x, y) (point (x, y )) is
Link 2
Ŷ0 ( )
∆ ẋ
l2 V=
θ2 X̂1 ẏ
[ ]( )
{0} −l1 s1 − l2 s12 −l2 s12 θ˙1
=
l1 c1 + l2 c12 l2 c12 θ˙2
O2
Link 1
l1
where θ̇1 , θ̇2 are joint rates.
θ1
The matrix in square brackets is the Jacobian
X̂0 matrix in {0}.
O1

. . . . . . . . . . . . . . . . . . . .
F IGURE : A planar 2R manipulator . . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 30 / 103
G EOMETRIC I NTERPRETATION OF
M ANIPULATOR JACOBIAN MATRIX

Magnitude of linear velocity vector

∆
|V|2 = V · V = g11 θ̇12 + 2g12 θ̇1 θ̇2 + g22 θ̇22 (23)

gij , i, j = 1, 2, are the elements of a matrix [ g ] = [J(Θ)]T [J(Θ)].

For the planar 2R manipulator the gij ’s are

g11 = l12 + l22 + 2l1 l2 c2

g12 = g21 = l22 + l1 l2 c2
g22 = l22 (24)

The elements gij ’s are functions of θ2 alone and g22 is a constant.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 31 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)
Maximum and minimum |V|2 subject to constraint θ̇12 + θ̇22 = 14 .
Solve ∂ |V∗ |2 /∂ θ̇i = 0, i = 1, 2, where

V∗ 2 = g11 θ̇12 + 2g12 θ̇1 θ̇2 + g22 θ̇22 − λ (θ̇12 + θ̇22 − 1)

Partial differentiation reduces to an eigenvalue problem

[ g ]Θ̇ − λ Θ̇ = 0 (25)

The eigenvalues are

λ1,2 = (1/2){(g11 + g22 ) ± [(g11 + g22 )2 − 4(g11 g22 − g12 )] }

2 1/2

4 Without any constraint V ∈ ℜ2 and fills up ℜ2 . The constraint θ̇ 2 + θ̇ 2 = 1 is similar to the

1 2. . . . . . . . . . . . . . . . . . . .
unit speed constraint in differential geometry of space curves. . . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 32 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)

[ g ] real, symmetric and positive definite → Eigenvalues are always real and
positive.
For λ1 > λ2 , √ √
|V|max = λ1 , λ2
|V|min =
√ √
For square Jacobian matrix, eigenvalues of [J(Θ)] are λ1 and λ2 (see
Strang 1976).
√ √
Maximum and minimum |V| for 2R manipulator are λ1 and λ2 .
If θ̇12 + θ̇22 = k 2 is used → Maximum and minimum |V| are scaled by k.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 33 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)
From V = [J(Θ)]Θ̇, [J]T V = [ g ]Θ̇
For non-singular [ g ],
T
VT ([J][ g ]−1 )([J][ g ]−1 )T V = Θ̇ Θ̇

For planar 2R manipulator, ([J][ g ]−1 )([J][ g ]−1 )T is symmetric and of rank 2
T
→ for Θ̇ Θ̇ = 1, (ẋ, ẏ )T ([J][ g ]−1 )([J][ g ]−1 )T (ẋ, ẏ ) = 1.
xT [A]x = 1, with [A] symmetric, positive definite, describes an ellipse.
The tip of the linear velocity vector
√ traces an
√ ellipse and the semi-major and
semi-minor axes of the ellipse are λ1 and λ2 , respectively.
T
For Θ̇ Θ̇ = k 2 , size of ellipse is scaled by k, but shape of ellipse does not
change with k.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 34 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)
Eigenvalues of [ g ] are only
functions of θ2 → Shape and size of
ellipse will change with θ2 .
(x, y)
Can plot ellipses at all points in the Ŷ0 Link 2
workspace l2
θ2
Recall: workspace of a planar 2R is {0}
the area between two circles of radii
O2
l1 + l2 and l1 − l2 . l1 Link 1

Ellipse independent of θ1 → All θ1

ellipses at a chosen radius (in the X̂0

O1
annular region) are same!
F IGURE : Velocity ellipse for a planar 2R
manipulator . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 35 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)

The shape of the velocity ellipse indicates which directions are ‘easier’ to move
for given joint rates
|V| is larger along major axis → Easier to move along major axis.
Less easier to move along the minor axis.
Ellipse reduces to a circle → Equally easy to move in all directions.
All points in the workspace, where the ellipse is a circle, are called isotropic (see
Salisbury, 1982)

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 36 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)
Isotropic configuration — Eigenvalues of [J(Θ)] (or [ g ]) are equal.
For planar 2R, eigenvalues equal only if

g11 = g22 and g12 = 0

From the expressions of gij ’s above conditions imply that

l12 + 2l1 l2 c2 = 0 and l22 + l1 l2 c2 = 0

√
and this is only possible if l1 = 2l2 and c2 = − √12
A planar 2R manipulator √ can posses isotropic configurations only if the link
lengths have a ratio of 2, and θ2 = 135◦ .
Since θ1 ∈ [0, 2π ], all the isotropic configurations lie on a circle.
Degenerate form of velocity ellipse → Singular configuration. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 37 / 103
G EOMETRIC I NTERPRETATION
(C ONTD .)

Spatial motion & 2 degree-of-freedom → Velocity vector on tangent plane to a

surface → Velocity ellipse.
Spatial motion & 3 degree-of-freedom → Velocity vector lies in ℜ3 → Velocity
ellipsoid.
Same ideas can be extended to angular velocity vector.
Extension to 6 × 6 manipulator Jacobian matrix – More complicated since not a
dimensionally homogeneous matrix.
Need to use notions of screws and twists (see Hunt, 1976).
Velocity ellipse → Cylindroid & Two screw system.
Velocity ellipsoid → Hyperboloid & Three screw system.
Extension to parallel manipulators using parallel manipulator Jacobian.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 38 / 103
N ON - SQUARE JACOBIAN M ATRIX

For square Jacobian → Matrix can be inverted to obtain joint rates.

Redundant systems Jacobian matrix is not square → Number of joint variables
more than 6 (for ℜ3 ) or more than 3 (for ℜ2 ).
Jacobian matrix cannot be inverted to obtain joint rates given linear and
angular velocity of end-effector.
Use of pseudo-inverse to resolve redundancy – Pseudo-inverse of m × n (n > m)
matrix [J(Θ)]
[J(Θ)]# = [J(Θ)]T ([J(Θ)][J(Θ)]T )−1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 39 / 103
N ON - SQUARE JACOBIAN
Some properties of pseudo-inverse
Dimension of [J(Θ)]# is n × m – Not square!
Left inverse → [J(Θ)][J(Θ)]# = [U] – Identity matrix.
Not a right inverse [J(Θ)]# [J(Θ)] ̸= [U]
General solution to 0 VTool =0Tool [J(Θ)]Θ̇ is

Θ̇ = [J(Θ)]# 0 VTool + ([U] − [J(Θ)]# [J(Θ)])W˙

([U] − [J(Θ)]# [J(Θ)])W˙ lies in the null-space of [J(Θ)].

T
Pseudo-inverse, without null-space, minimises Θ̇ Θ̇.
The null-space term used to avoid obstacles, joint limits and to maximise a
manipulability index det([[J(Θ)][J(Θ)]T ]1/2 )(see Nakamura, 1991).
Disadvantages
Local numerical scheme – No global or analytical results.
Velocity level and not a position and orientation level scheme for resolution of
redundancy. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 40 / 103
S UMMARY
Propagation of velocities can be used to obtain linear and angular velocity of
end-effector in terms of joint rates.
Manipulator Jacobian matrix relate end-effector linear and angular velocity to
joint rates.
Manipulator Jacobian matrix is not dimensionally homogeneous!
Geomtric interpretation – Velocity ellipse and ellipsoid → Ease of motion
direction.
Geometric interpretation of manipulator Jacobian for linear and angular velocity
can be done separately.
More complicated description in terms of screw cylindroid etc. when full rigid
body motion is considered!
Non-square Jacobian matrix –>pseudo-inverse –> Resolution of redundancy at
velocity level.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 41 / 103
O UTLINE
1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 42 / 103
PARALLEL M ANIPULATOR JACOBIAN
M ATRIX
Parallel manipulators has both actuated and passive joints — q = (θ , ϕ )T
Loop-closure equations do not contain all joint variables.
No natural choice of end-effector {Tool} → No velocity propagation.
Platform type parallel manipulator – Position of centroid & orientation of
platform {Tool} is of interest.
Linear and angular velocity of centroid and {Tool}
d 0 T
0
ω Tool = ( [R]) 0Tool [R] =0Tool [Jω (q)] q̇
dt Tool
1 0˙
0
VTool = ( p1 + 0 p˙ 2 + 0 p˙ 3 ) =0Tool [JV (q)] q̇ (26)
3

0 0
Tool [Jω (q)], Tool [JV (q)]
– Angular and linear velocity Jacobian.
q̇ – Time derivatives of configuration variables q.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 43 / 103
E LIMINATION OF PASSIVE JOINT
RATES

Linear and angular velocity function of all q and q̇.

Only the actuated joints θi , i = 1, 2, ..., n are specified.
The m passive ϕi ’s can be obtained from direct kinematics
Need expression for ϕ̇i and obtain linear and angular velocities in terms of only
θ̇i ’s.
Derived from the m loop-closure or constraint equations.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 44 / 103
E LIMINATION OF PASSIVE JOINT
RATES
For m passive variables, m constraint equations ηi (q1 , ..., qn+m ) = 0, i = 1, ..., m
or in a vector form
η (q) = η (θ , ϕ ) = 0 (27)
Differentiate equation (27) with respect to t, and rearrange

[K (q)]θ̇ + [K ∗ (q)]ϕ̇ = 0 (28)

Columns of the m × n matrix [K (q)] are the partial derivatives of η (q) with
respect to the actuated variables θi , i = 1, ..., n,
Columns of m × m matrix [K ∗ (q)] are the partial derivatives of η (q) with respect
to the passive variables ϕi , i = 1, ..., m.
[K ∗ (q)] is always an m × m square matrix.
[K (q)] and [K ∗ (q)] are functions q = (θ , ϕ ) ∈ ℜn+m . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 45 / 103
E LIMINATION OF PASSIVE JOINT
RATES
If det([K ∗ ]) ̸= 0,
ϕ̇ = −[K ∗ ]−1 [K ]θ̇ (29)
The angular and linear velocity can partitioned as
∗
0
ω Tool = [Jω ]θ̇ + [Jω ]ϕ̇ , 0
VTool = [JV ]θ̇ + [JV∗ ]ϕ̇

Substitute ϕ̇ to get
∗
0
ω Tool = ([Jω ] − [Jω ][K ∗ ]−1 [K ])θ̇ , 0
VTool = ([JV ] − [JV∗ ][K ∗ ]−1 [K ])θ̇

Define equivalent [Jω ]eq and [JV ]eq

[JV ]eq = [JV ] − [JV∗ ][K ∗ ]−1 [K ]

∆
(30)
∗
][K ∗ ]−1 [K ]
∆
[Jω ]eq = [Jω ] − [Jω .
(31).
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 46 / 103
E QUIVALENT JACOBIAN MATRIX IN
PARALLEL MANIPULATORS

Using [JV ]eq and [Jω ]eq

 0V

Tool
VTool =  −−  =
∆
Tool [Jeq ]θ̇
0 0
(32)
0ω
Tool

The 6 × n matrix, 0
Tool [Jeq ], consists of 3 × n rows from [JV ]eq and 3 × n rows
from [Jω ]eq .
The matrix 0 is the Jacobian matrix5 for parallel manipulators.
Tool [Jeq ]
At a known q, equation (32) relate actuated joint rates θ̇ to the linear and
angular velocity of chosen end-effector {Tool}.

5 Not . . . . . . . . . . . . . . . . . . . .
a “true Jacobian” as it is not obtained from the derivative of a vector
. . . valued
. . . . . function.
. . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 47 / 103
E QUIVALENT JACOBIAN MATRIX IN
PARALLEL MANIPULATORS

The matrix 0 can be used define a [ gV ]eq for parallel manipulators.

Tool [JV ]eq

[ gV ]eq = ([JV ] − [JV ∗ ][K ∗ ]−1 [K ])T ([JV ] − [JV ∗ ][K ∗ ]−1 [K ]) (33)

[ gV ]eq is symmetric and positive definite.

Similar to a serial manipulator, the tip of the linear velocity vector lies on an
ellipse or an ellipsoid.
Much more complicated than in serial manipulators!
[ gω ]eq defined using [Jω ]eq → Angular velocity ellipse or ellipsoid.
The above geometrical description is valid if det [K ∗ ] ̸= 0.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 48 / 103
E XAMPLE – P LANAR 4- BAR
φ3
MECHANISM
O2
Link 2
X̂2
O3
l2
X̂3

φ2
Ŷ L
l3 ŶR
Link 1 Link 3
{L} l1
{R}
X̂1
φ1
θ1 X̂L
l0
X̂R
OT ool , OR
OL , O1
X̂T ool

F IGURE : The four-bar mechanism

“Break” at third joint – a planar 2R + a planar 1R manipulator

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 49 / 103
E XAMPLE – P LANAR 4- BAR
MECHANISM

Constraint equation of a four-bar

∆
η1 (q) = l1 cos θ1 + l2 cos(θ1 + ϕ2 ) − l0 − l3 cos ϕ1 = 0
∆
η2 (q) = l1 sin θ1 + l2 sin(θ1 + ϕ2 ) − l3 sin ϕ1 = 0

θ1 is the actuated joint variable and (ϕ1 , ϕ2 ) are the passive joint variables.
Derivative of constraint equations with respect to time t gives
( )
−l1 sin θ1 − l2 sin(θ1 + ϕ2 )
θ˙1 +
l1 cos θ1 + l2 cos(θ1 + ϕ2 )
( )( )
l3 sin ϕ1 −l2 sin(θ1 + ϕ2 ) ϕ˙1
=0
−l3 cos ϕ1 l2 cos(θ1 + ϕ2 ) ϕ˙2
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 50 / 103
E XAMPLE – P LANAR 4- BAR
MECHANISM

[K ] and [K ∗ ] matrices for the planar 4-bar are

( )
−l1 sin θ1 − l2 sin(θ1 + ϕ2 )
[K ] =
l1 cos θ1 + l2 cos(θ1 + ϕ2 )
[ ]
∗ l3 sin ϕ1 −l2 sin(θ1 + ϕ2 )
[K ] =
−l3 cos ϕ1 l2 cos(θ1 + ϕ2 )
The matrix [K ∗ ] is a square 2 × 2 matrix.
[K ] and [K ∗ ] matrices are functions of the actuated and passive variables.
Fairly simple to calculate for planar 4-bar.
Multi-degree-of-freedom spatial mechanisms → Use symbolic algebra software
such as MAPLE⃝ R
.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 51 / 103
E XAMPLE – 3-RPS M ANIPULATOR
p(x, y, z) Moving Platform
Ẑ

S3 S2

S1 l2

l3 θ2 θi , i = 1, 2, 3 are passive variables.

l1
Axis of R2 li , i = 1, 2, 3 are actuated variables.
θ3 {Base}
“Break” at the spherical joints
O
Obtain position vector of spherical
Ŷ
joints Si with respect to O in {Base}
θ1
Use S − S pair constraints
Base Platform
Axis of R3

Axis of R1
X̂

F IGURE : The 3-RPS parallel manipulator . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 52 / 103
E XAMPLE – 3-RPS PARALLEL
MANIPULATOR

For the 3-RPS manipulator, 3 constraint equations are

η1 (q) = 3 − 3a2 + l12 + l22 + l1 l2 c1 c2 − 2l1 l2 s1 s2 − 3l1 c1 − 3l2 c2 = 0
η2 (q) = 3 − 3a + l22 + l32 + l2 l3 c2 c3 − 2l2 l3 s2 s3 − 3l2 c2 − 3l3 c3
2
= 0
η3 (q) = 3 − 3a2 + l32 + l12 + l3 l1 c3 c1 − 2l3 l1 s3 s1 − 3l3 c3 − 3l1 c1 = 0

Perform the derivative of ηi (q), i = 1, 2, 3, with respect to time and rearrange

to obtain [K ] and [K ∗ ]
[K ] involves derivative with respect to the actuated variables l1 , l2 and l3
 
2l1 − 3c1 + l2 c1 c2 − 2l2 s1 s2 2l2 − 3c2 + l1 c1 c2 − 2l1 s1 s2 0
 0 2l2 − 3c2 + l3 c2 c3 − 2l3 s2 s3 2l3 − 3c3 + l2 c2 c3 − 2l2 s2 s3 
2l1 − 3c1 + l3 c1 c3 − 2l3 s1 s3 0 2l3 − 3c3 + l1 c1 c3 − 2l1 s1 s3

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 53 / 103
E XAMPLE – 3-RPS PARALLEL
MANIPULATOR
[K ∗ ] involves derivative with respect to passive joint variables, θ
 
3l1 s1 − l1 l2 s1 c2 − 2l1 l2 c1 s2 3l2 s2 − l1 l2 c1 s2 − 2l1 l2 s1 c2 0
 0 3l2 s2 − l2 l3 s2 c3 − 2l2 l3 c2 s3 3l3 s3 − l2 l3 c2 s3 − 2l2 l3 s2 c3 
3l1 s1 − l1 l3 s1 c3 − 2l1 l3 c1 s3 0 3l3 s3 − l1 l3 c1 s3 − 2l1 l3 s1 c3

For the centroid, [JV ] and [JV ∗ ], are

 
−c1 (1/2)c
√ 2 (1/2)c3
√
[JV ] = (1/3)  0 (− 3/2)c2 ( 3/2)c3 
s1 s2 s3

and  
l1 s1 −(1/2)l
√ 2 s2 (−1/2)l
√ 3 s3
[JV ∗ ] = (1/3)  0 ( 3/2)l2 s2 (− 3/2)l3 s3 
l 1 c1 l 2 c2 l 3 c3 . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 54 / 103
E XAMPLE – 3-RPS PARALLEL
MANIPULATOR

To obtain [Jω ] and [Jω ∗ ], compute dt

d Base T
(Top [R]) Base
Top [R] and then rearrange.
Expressions are too large — For l1 = 2/3, l2 = 3/5, l3 = 3/4 and corresponding
θ1 = 0.7593, θ2 = 0.2851, θ3 = 0.8028 radians,
   
0.0644 −0.2801 −0.9548 −0.0307 0.6686 −0.3398
[Jω ] =  −1.1519 0.1361 0.4815  , [Jω ] =  −0.3961
∗
0.4256 0.1713 
−0.1953 0.5339 −0.3803 0.3069 0.0308 −0.1353

Expressions for [JV ]eq and [Jω ]eq are more harder to obtain as [K ∗ ]−1 is
needed. For above numerical values
   
−0.2313 0.5372 0.0114 2.1409 −6.4331 0.4665
[JV ]eq =  0.0722 −0.6758 0.1951  , [Jω ]eq =  0.0072 −4.1216 1.6048 
1.1765 −1.6830 0.9223 0.1565 0.4570 −0.3285

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 55 / 103
E XAMPLE – 3-RPS PARALLEL
MANIPULATOR

For a = 1/2, and (l1 , l2 , l3 ) = (0.5, 1.0, 2.0) meters →

(θ1 , θ2 , θ3 ) = (0.4, 0.7535, 0.2402) radians by direct kinematics.
Tip of linear velocity vector of centroid lies on an ellipsoid – Shown in Figure as
three sectional views and a 3D plot.
Maximum, intermediate, and minimum velocities along the principal axes of the
ellipsoid are 0.3724, 0.3162, 0.2031 m/sec, respectively.
The directions of principal axes are (0.9921, −0.0394, 0.1187)T ,
(0.1166, 0.6338, −0.7646)T and (−0.0452, 0.7724, 0.6335)T , respectively.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 56 / 103
E XAMPLE – 3-RPS MANIPULATOR
0.4 0.4

0.2 0.2

Vy

Vz
0 0

−0.2 −0.2

−0.4 −0.4
−0.4 −0.2 0 0.2 0.4 −0.4 −0.2 0 0.2 0.4
Vx Vx

0.4

0.5
0.2

Vz
0
Vz

−0.2 −0.5
0.5
0.5
0 0
−0.4
−0.4 −0.2 0 0.2 0.4 Vy −0.5 −0.5 Vx
Vy

F IGURE : Velocity ellipsoid at a non-singular point

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 57 / 103
S UMMARY

Parallel manipulator Jacobian in terms of active and passive joint variables.

Two more matrices, [K ] and [K ∗ ], arise from derivative of constraint equations.
Can solve for passive joint rates ϕ̇i and obtain equivalent Jacobian matrix.
Can obtain equivalent Jacobian only if det[K ∗ ] ̸= 0.
Can obtain geometric interpretation as in serial manipulators – Ellipse and
ellipsoids.
More difficult to obtain numerical results – Elimination of passive variables!

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 58 / 103
O UTLINE
1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 59 / 103
S ERIAL MANIPULATORS – R EVIEW
Direct velocity kinematics
0
VTool = 0Tool [J(Θ)]Θ̇

For known Θ and Θ̇, linear and angular velocity of end-effector obtained from
above equation.
0
VTool always exists
Inverse velocity kinematics
−1 0
Θ̇ = 0Tool [J(Θ)] VTool

Joint rates can be obtained when Jacobian matrix is square, and

det(0Tool [J(Θ)]) ̸= 0
det(0Tool [J(Θ)]) = 0 → Loss of rank of 0Tool [J(Θ)] → Singular configuration.
At singular configuration, Θ̇ cannot be obtained for given linear and angular
velocity. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 60 / 103
S INGULARITY IN PLANAR 2R
MANIPULATOR

For a planar 2R manipulator

( ) [ ]( )
ẋ −l1 s1 − l2 s12 −l2 s12 θ˙1
=
ẏ l1 c1 + l2 c12 l2 c12 θ˙2

The Jacobian matrix is

[ ]
0 −l1 s1 − l2 s12 −l2 s12
Tool [J(Θ)] =
l1 c1 + l2 c12 l2 c12

det(0Tool [J(Θ)]) = 0 → sin θ2 = 0.

This implies θ2 = 0, π → Second link is stretched completely or folded on top
of first link.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 61 / 103
 S INGULARITY IN PLANAR 2R
MANIPULATOR
Possible directions radius = l1 + l2
of velocity
vector
Planar 2R manipulator for θ2 = 0, π .
Ŷ0 End-effector can only move
l2
perpendicular to the line O1 − O2
{0} Link 2
l1
connecting the two rotary joints.
O2
Link 1
The end-effector cannot have a
θ1
velocity component along the second
O1
X̂0
link.
Instantaneous loss (only at this
configuration) of one degree of
radius = l1 − l2
freedom.
F IGURE : Singular configurations for a planar
2R manipulator (l1 > l2 ) . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 62 / 103
S INGULARITY IN PLANAR 2R
MANIPULATOR
At θ2 = 0, π , the velocity ellipse degenerates to a line — Along the possible
direction of motion.
For the 2R planar example, the Jacobian matrix can be inverted easily, and
( ) ( )( )
θ̇1 1 l2 c12 l2 s12 ẋ
=
θ̇2 l1 l2 s2 −l1 c1 − l2 c12 −l1 s1 − l2 s12 ẏ

As θ2 → 0 or π , s2 → 0, and (θ̇1 , θ̇2 )T → ∞.

Knowledge of singularity is important – When det(0Tool [J(Θ)]) is close to zero,
joint velocities tend to become large and cause problems for controller of robot.
Singularities occurs in all serial manipulator and not only in planar 2R.
For planar 2R, singularity only at workspace boundaries. In other manipulators
can happen elsewhere also! . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 63 / 103
PARALLEL MANIPULATORS – R E -
VIEW
In parallel manipulators, equivalent Jacobian need to be used.
For parallel manipulators, the linear and angular velocity Jacobians are

[JV ]eq = [JV ] − [JV∗ ][K ∗ ]−1 [K ]

∆

∗
][K ∗ ]−1 [K ]
∆
[Jω ]eq = [Jω ] − [Jω
Equivalent Jacobian  
[JV ]eq
0
Tool [J]eq =
 −− 
[Jω ]eq
det(0Tool [Jeq ]) = 0 → End-effector loses one or more degrees of freedom →
Actuated joint rates Θ̇ → ∞.
Similar to serial manipulators — At singularity velocity ellipse or ellipsoid
degenerates. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 64 / 103
PARALLEL MANIPULATORS – G AIN
SINGULARITY

In addition to loss singularity, there exists a second kind of singularity in parallel

manipulators.
θ̇ = 0 → Actuated joints locked → Mechanism becomes a structure.
Equation [K (q)]θ̇ + [K ∗ (q)]ϕ̇ = 0 becomes

[K ∗ (q)]ϕ̇ = 0
∗
From linear algebra, non-zero solution ϕ̇ exists when det([K ∗ ]) = 0.
∗
ϕ̇ is the eigenvector corresponding to the zero eigenvalue of [K ∗ ].

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 65 / 103
PARALLEL MANIPULATORS – G AIN
SINGULARITY

∗
For a non-zero ϕ̇ , and θ̇ = 0,
∗
0
ω Tool = [Jω ∗ ]ϕ̇
∗
0
VTool = [JV ∗ ]ϕ̇

Even with actuators locked the linear and angular velocity are non-zero.
The end-effector of the parallel manipulator can instantaneously gain one or
more degrees of freedom.
Termed as gain singularity.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 66 / 103
PARALLEL MANIPULATORS – G AIN
SINGULARITY

Geometric picture: Non-singular configuration.

At non-singular configurations, θ̇ = 0 → ϕ̇ = 0 → 0 VTool = 0.
At a non-singular position velocity ellipsoid is of zero size.
Geometric picture: Gain singularity configuration
Loss of rank of [K ∗ ].
If rank is (m − 1) → There exists non zero eigenvector ϕ̇ 1 for the zero eigenvalue
of [K ∗ ].
C1 ϕ̇ also an eigenvector with C1 a scaling constant.
For θ̇ = 0
0
VTool = C1 [JV ∗ ]ϕ̇ 1

There can be motion along [JV ∗ ]ϕ̇ 1 !

The zero velocity ellipsoid ‘grows’ into a line.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 67 / 103
PARALLEL MANIPULATORS – G AIN
SINGULARITY

If rank of matrix [K ∗ ] is (m − 2), then

0
VTool = C1 [JV ∗ ]ϕ̇ 1 + C2 [JV ∗ ]ϕ̇ 2

ϕ̇ 1 , ϕ̇ 2 are eigenvectors from the two zero eigenvalues of [K ∗ ].

C1 , C2 are the two scaling constants.
For C12 + C22 = 1, tip of velocity vector traces an ellipse6 .
If rank of [K ∗ ] is (m − 3), then tip of velocity vector will lie on an ellipsoid.
If rank is less than (m − 3) and only 0 VTool is of interest → Similar to a
redundant serial manipulator.
6 C and C are similar to θ̇ and θ̇ and C 2 + C 2 = 1 is similar to the constraint θ̇ 2 + θ̇ 2 = 1
1 2 1 2 1 2 1 2
used in the planar 2R example. Using same reasoning as in 2R case, the tip of 0 VTool for a parallel
. . . . . . . . . . . . . . . . . . . .
manipulator lies on an ellipse. . . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 68 / 103
G AIN SINGULARITY

Gain singularity occurs in parallel/hybrid manipulators.

In fully-parallel six DOF manipulators (end-effector directly connected to base
by one actuated joint – Stewart-Gough platform) only gain singularity possible
(Hunt, 1991).
In six DOF hybrid/parallel manipulator (example three-fingered hand), both
loss and gain singularity possible.
Gain singularity is related to capability of resisting external force or moments.
Large amount of literature on singularity analysis of parallel manipulators (for
example, Hunt (1986), Litvin et. al (1990), Merlet (1991), Gosselin and
Angeles (1990), Zlatanov (1995), Park and Kim (1999)).
Singularity analysis & uses of singularity is an active topic of research (see
Bandyopadhyay and Ghosal (2009), Ranganath et al. (2004)).
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 69 / 103
G AIN S INGULARITY (C ONTD .)
Special link lengths and geometry → Gain over finite range of motion.
Passive link Ocan show instantaneous and finite dwell.
3

l2
l3
Link 2
Link 3

Link 2 and link 3 can rotate from

0 to 2π with θ1 and θ2 locked
Ŷ L O2
ŶR
{L}
l1
Link 4
{R} (see Bandyopadhyay and Ghosal
Link 1 l4
(2004a) for details).
θ2
θ1 X̂L
l0
OR X̂R
OL

F IGURE : Finite motion at gain singularity

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 70 / 103
S INGULARITIES - 4- BAR
O2

φ2

Link 2
Ŷ L l2
ŶR
Link 1
{L} l1 O3
{R}
l3
Link 3 φ1
θ1 X̂L
l0 X̂R

OR
OL , O1

F IGURE : Singular configuration for a planar four-bar mechanism

det([K ∗ ]) = 0 → l2 l3 sin(θ1 + ϕ2 − ϕ1 ) = 0 → θ1 + ϕ2 − ϕ1 = nπ → ϕ3 = 2π ,
Link 2 and 3 are parallel.
Instantaneous gain: θ1 locked, point O2 is fixed → Link 2 and link 3 along a
straight line → O3 can have instantaneous velocity along the common tangent..
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 71 / 103
S INGULARITIES IN PARALLEL
MANIPULATORS – E XAMPLES

Example – The 3-RPS parallel manipulator (see Basu and Ghosal, 1997)
For the 3-RPS parallel manipulator,
det([JV ]eq ) = 0 → Linear velocity ellipsoid described by the centroid of the top
platform degenerates to an ellipse.
For (l1 , l2 , l3 ) = (0.5, 1.0, 1.9710) meters and
(θ1 , θ2 , θ3 ) = (1.1691, 0.4781, 0.2355) radians → det([JV ]eq ) = 0.
The linear velocity ellipse at this configuration is shown in sectional and a 3D
view in Figure.
Not a contradiction to result by Hunt (1991)
The 3-RPS parallel manipulator is not a six DOF manipulator, and
Only the linear velocity vector of the centroid is considered.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 72 / 103
E XAMPLE – 3-RPS MANIPULATOR
2 2

1 1

Vy

Vz
0 0

−1 −1

−2 −2
−2 −1 0 1 2 −2 −1 0 1 2
Vx Vx

2
1

Vz
0
Vz

−1 −2
2
2
0 0
−2
−2 −1 0 1 2 Vy −2 −2 Vx
Vy

F IGURE : Linear velocity ellipse at a loss singular point

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 73 / 103
S INGULARITIES IN PARALLEL
MANIPULATORS – E XAMPLES
Example – 3-RPS parallel manipulator
Gain one or more degrees-of-freedom when det([K ∗ ]) = 0 i.e.,

det([K ∗ ]) =
(3l1 s1 − l1 l2 s1 c2 − 2l1 l2 c1 s2 ) × (3l2 s2 − l2 l3 s2 c3 − 2l2 l3 c2 s3 ) ×
(3l3 s3 − l1 l3 c1 s3 − 2l1 l3 s1 c3 )
+(3l1 s1 − l1 l3 s1 c3 − 2l1 l3 c1 s3 ) × (3l2 s2 − l1 l2 c1 s2 − 2l1 l2 s1 c2 ) ×
(3l3 s3 − l2 l3 c2 s3 − 2l2 l3 s2 c3 ) = 0

det([K ∗ ]) is a function of all (θ , ϕ )

det([K ∗ ]) = 0 and three loop-closure equations → Four equations in six
variables → A 2D surface.
Difficult to eliminate and get analytical expression. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 74 / 103
E XAMPLE - -3-RPS MANIPULATOR

For (l1 , l2 , l3 ) = (0.575, 0.483, 0.544), and

(θ1 , θ2 , θ3 ) = (−0.3441, −0.0138, 0.2320) radians, det[K ∗ ] ≈ 0.
The eigenvalues of [K ∗ ] are approximately −0.5565, 0 and 0.4509.
The three corresponding eigenvectors are (−0.8098, 0.3571, −0.4656)T ,
(−0.3109, −0.8743, −0.3727)T and (−0.0877, −0.4781, −0.8739)T .
Gained velocity of centroid is
     
−0.0647 0.0011 −0.0208
0
VTool = 0  θ̇1 +  −0.0019  θ̇2 +  −0.0361  θ̇3
0.1804 0.1610 0.1763

(θ̇1 , θ̇2 , θ̇3 )T = α (−0.3109, −0.8743, −0.3727)T with α arbitrary.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 75 / 103
E XAMPLE – 3-RPS MANIPULATOR
0.05 1

0.5

Vy

Vz
0 0

−0.5

−0.05 −1
−0.1 −0.05 0 0.05 0.1 −0.1 −0.05 0 0.05 0.1
Vx Vx

1
0.5

Vz
0
Vz

−0.5 −1
0.05
0.1
0 0
−1
−0.05 0 0.05 Vy −0.05 −0.1 Vx
Vy

F IGURE : Velocity at a gain singular point

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 76 / 103
S INGULARITIES IN PARALLEL
MANIPULATORS – E XAMPLES

Example – 3-RPS parallel manipulator

At (l1 , l2 , l3 ) = (1.9363, 2.9998, 1.9363) meters
Corresponding (θ1 , θ2 , θ3 ) = (1.3096, 0.9817, 1.3096) radians,
det[K ∗ ] ≈ 0 → Eigenvalues are approximately 0, 0, 3.9680.
At this configuration, gain of two degrees of freedom.
The singularities corresponding to gain of two degrees of freedom lie on a curve
in ℜ3 → Difficult to get analytical expression.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 77 / 103
E XAMPLES - 3-RPS MANIPULATOR
1.5 1

1
0.5
0.5

Vy

Vz
0 0

−0.5
−0.5
−1

−1.5 −1
−0.4 −0.2 0 0.2 0.4 −0.4 −0.2 0 0.2 0.4
Vx Vx

1
0.5

Vz
0
Vz

−0.5 −1
2
0.5
0 0
−1
−2 −1 0 1 2 Vy −2 −0.5 Vx
Vy

F IGURE : Velocity ellipse at a gain singular point

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 78 / 103
S UMMARY

Singularity in serial manipulators

Loss of rank of manipulator Jacobian → Loss of one or more DOF.
Velocity ellipsoid/ellipse degenerates.
Not possible to move along singular direction(s).
Singularity in parallel manipulators
Loss singularity – Loss of DOF as in serial manipulators.
Can also gain one or more DOF with actuators locked.
Gain of DOF due to loss of rank of [K ∗ ] and det[K ∗ ] = 0
Examples of a planar serial 2R and a 3-RPS parallel manipulator.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 79 / 103
O UTLINE
1 C ONTENTS
2 L ECTURE 1
Introduction
Linear and Angular Velocity of Links
3 L ECTURE 2
Serial Manipulator Jacobian Matrix
4 L ECTURE 3
Parallel Manipulator Jacobian Matrix
5 L ECTURE 4
Singularities in Serial and Parallel Manipulators
6 L ECTURE 5
Statics of Serial and Parallel Manipulators
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 80 / 103
S TATICS OF S ERIAL AND PARALLEL
M ANIPULATORS

Joints of a serial manipulator are locked → Manipulator becomes a structure.

Forces and moments acting at joints when manipulator structure is subjected to
external forces and moments.
External forces and moments on end-effector if pushing some object or carrying
a payload.
Useful to know joint forces or torques which can maintain the static equilibrium.
Use free-body diagram and equations of static equilibrium.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 81 / 103
 S ERIAL MANIPULATORS – S TATICS
ni+1 {i + 1} Ẑi+1 fi+1

Ŷi+1
Ẑi
{i} i
Oi+1

Oi+1 X̂i+1

ni
Two intermediate rotary (R) joints
and a link of a manipulator.
Ŷi Link i
Oi fi and ni denote the forces and
moments exerted on link {i} by link
X̂i {i − 1}.
fi

F IGURE : Free-body diagram of a link

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 82 / 103
S ERIAL MANIPULATORS – S TATICS

For static equilibrium of {i}, ΣF = 0

i
fi −i fi+1 = 0

fi +1 is the force on link {i + 1} exerted by link {i} → Force on link {i} exerted
by link {i + 1} will be equal and of opposite sign.
The leading superscript i signifies that the vectors are described in {i}.

For static equilibrium of {i}, ΣM = 0

i
ni −i ni+1 −i Oi+1 ×i fi+1 = 0

iO is the vector from Oi to Oi +1 .

i +1
Negative sign due to same reason as for forces.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 83 / 103
S ERIAL MANIPULATORS – S TATICS
Equilibrium equations can be written as
i i i+1
fi = i+1 [R] fi+1
i
ni = i
i+1 [R]
i+1
ni+1 +i Oi+1 ×i fi (34)
Inward recursion for forces and moments on each link.
Forces and moments at the end-effector: n+1 fn+1 =n+1 nn+1 = 0 if not in contact
with environment.
n+1 f n+1 n
n+1 , n+1 known otherwise.
Recursively compute i fi , i ni for i : n to 1 using equation (34).
Joint can only apply force or moment along Ẑ − axis; all other components
resisted by structure/bearings.
Torque required at joint i to maintain equilibrium
τi = i
ni ·i Ẑi (joint i is rotary (R))
τi = i
fi · Ẑi
i
(joint i is prismatic (P)) (35)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 84 / 103
S TATICS – 3R MANIPULATOR n

X̂3 , X̂T ool

3R planar manipulator applying force Link 3

{T ool} lf 3 X̂2
θ3
0 f
fTool = (fx , fy , 0)T ŶT ool
τ3 O3
0
nTool = (0, 0, nz )T Link 2
l2
Ŷ0 τ2
In {Tool} coordinate system θ2 X̂1
    
fx′ c123 s123 0 fx {0}

 fy  =  −s123
′ c123 0   fy  l1 O2
τ1 Link 1
0 0 0 1 0
θ1

and (0, 0, nz′ )T = (0, 0, nz )T . X̂0

O1

F IGURE : A 3R manipulator applying

force and moment
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 85 / 103
S TATICS – E XAMPLE 3R PLANAR
MANIPULATOR
i=3
3
f3 = (fx′ , fy′ , 0)T
3
n3 = (0, 0, nz′ + l3 fy′ )T

i=2
2
f2 = (c3 fx′ − s3 fy′ , s3 fx′ + c3 fy′ , 0)T
2
n2 = (0, 0, nz′ + l2 (s3 fx′ + c3 fy′ ) + l3 fy′ )T

i=1
1
f1 = (c23 fx′ − s23 fy′ , s23 fx′ + c23 fy′ , 0)T
1
n1 = (0, 0, nz′ + l1 (s23 fx′ + c23 fy′ ) + l2 (s3 fx′ + c3 fy′ ) + l3 fy′ )T
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 86 / 103
S TATICS – E XAMPLE 3R PLANAR
MANIPULATOR

Finally, the joint torques required to maintain equilibrium

τ1 =1 n1 ·1 Ẑ1 = nz′ + fx′ (l1 s23 + l2 s3 ) + fy′ (l1 c23 + l2 c3 + l3 )

τ2 =2 n2 ·2 Ẑ2 = nz′ + fx′ l2 s3 + fy′ (l2 c3 + l3 )
τ3 =3 n3 ·3 Ẑ3 = nz′ + fy′ l3

Above equations can be re-arranged as

 
fx
  fy 
−l1 s1 − l2 s12 − l3 s123 l1 c1 + l2 c12 + l3 c123 0 0 0 1 


0
τ = −l2 s12 − l3 s123 l2 c12 + l3 c123 0 0 0 1 


 (36)
0
−l2 s123 l3 c123 0 0 0 1 


0
nz
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 87 / 103
F ORCE TRANSFORMATION MATRIX

Term in the square bracket is the transpose of the Jacobian matrix.

As in velocities, denote forces and moments acting on the end-effector by
 0 
fTool
FTool =  −−  = (fx fy fz ; nx ny nz )T
0 ∆
(37)
0n
Tool

Note: 0 FTool is not a 6 × 1 vector since forces and moments have different
units.
0F is called a wrench in theoretical kinematics, and a wrench can be
Tool
thought of as screw with a magnitude which has units of force.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 88 / 103
F ORCE TRANSFORMATION MATRIX
Consider an infinitesimal Cartesian displacement of the end effector 0 δ X Tool 7
and the virtual work done by 0 FTool
Equating virtual work done by external force/moment and at joints

∆0
0
FTool ·0 δ X Tool = fTool · δ x +0 nTool · δ θ = τ · δ Θ

From definition of Jacobian, 0 δ X Tool = 0Tool [J(Θ)]δ Θ,

0
FTool · 0Tool [J(Θ)]δ Θ = τ · δ Θ

Above equations hold true for all δ Θ, hence

T 0
τ = 0Tool [J(Θ)] FTool (38)

Not surprising transpose of Jacobian appears in statics!

70 δ X . . . . . . . . . . . . . . . . . . . .
is 6 × 1 is a twist. The infinitesimal change in position and orientation . . . (.δ .x;. δ. θ
T
Tool . . . . . . . are . ). .. . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 89 / 103
PARALLEL MANIPULATORS –
S TATICS
T
For serial manipulators τ = 0Tool [J(Θ)] 0 FTool .
Principle of virtual work equally applicable for parallel manipulator
T 0
τ = 0Tool [Jeq ] FTool

{Tool} is a chosen end-effector.

0
Tool [Jeq ] is the equivalent Jacobian – Function of q, and
τ is the vector of forces or torques applied at the actuated joints only.
Difficult to compute 0Tool [Jeq ] due to [K ∗ ]−1 .
Inverse problem: Obtaining forces/moments applied by {Tool}
−T
0
FTool = 0Tool [J(q)eq ] τ
Inverse of Jacobian even more difficult!. Simpler approach for Gough-Stewart
platform. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 90 / 103
 PARALLEL MANIPULATORS –
S TATICS
{P0 }
Pi
P0
pi
S Joint

P Joint

The vector along the leg, B0 S

i,
B0
t
li
Ẑ B0
Si = B
P0 [R] pi + t − bi
0 P0 B0 B0
Ŷ

Bi B0 S
Unit vector along leg B0 s = i
i li
B0
bi
X̂

{B0 }
U Joint

F IGURE : A leg of a Stewart-Gough platform . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 91 / 103
PARALLEL MANIPULATORS –
S TATICS

The force exerted by actuated prismatic joint is fi B0 si .

The moment of the force (about the origin B0 ) fi (B0 bi ×B0 si ).
Denoting external force and moment by 0 FTool ,
 B   
0F
Tool ∑6i=1 B0 si fi
FTool =  − − −  =  
∆
B0
−−−
B0 M
Tool ∑i=1 ( bi × si )fi
6 B 0 B 0

where B0 FTool , B0 M
Tool are the 3 × 1 force and moment vectors acting on
{Tool}.
In matrix form B0 F
Tool =B0
Tool [ H ]f

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 92 / 103
F ORCE TRANSFORMATION MATRIX

B0
The force transformation matrix Tool [ H ]
 B0 s B0 s B0 s

1 2 ... 6
B0
Tool [ H ]= −−− −−− −−− −−−  (39)
(B0 b1 ×B0 s1 ) (B0 b2 ×B0 s2 ) ... (B0 b6 ×B0 s6 )

f = (f1 , f2 , ..., f6 )T is the vector of forces applied at the prismatic joints.

B0
Like the manipulator Jacobian matrix, Tool [ H ] is not a dimensionally
homogeneous matrix.
B0 −T
Tool [ H ] = 0Tool [J(q)eq ] but 0
Tool [ H ] is much simpler to compute!
Easily extended to any fully parallel manipulator – there are n columns.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 93 / 103
PARALLEL MANIPULATORS –
S INGULARITY IN FORCE DOMAIN

Direct force analysis: Obtain external force/moment given leg forces

B0
FTool = B 0
Tool [ H ]f

Inverse force analysis: Obtain leg forces given external force/moment

−1 B
f =B 0
Tool [ H ]
0
FTool

If det([ H ]) = 0, then inverse problem cannot be solved → Force singularity.

As det([ H ]) → 0, f → ∞, and any external force/moment along certain
direction cannot be resisted by the parallel manipulator8 .

8 In . . . . . . . . . . . . . . . . . . . .
velocity singularity, no joint rates can cause motion along certain (singular)
. . . . . . directions.
. . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 94 / 103
PARALLEL MANIPULATORS –
S INGULARITY IN FORCE DOMAIN

Force singularity can be visualised using degeneracy of force ellipsoid.

In a Stewart-Gough platform, external force F is given by
[ ]
F = [HF ]f = s1 s2 s3 s4 s5 s6 f

|F|2 is FT F = f T [gF ]f where [gF ] = [HF ]T [HF ].

The maximum, intermediate and minimum values of FT F subject to a
constraint of the form f T f = 1 are the eigenvalues of [gF ].
Since the rank of [gF ] is 3, the tip of the force vector F lies on an ellipsoid in
ℜ3 .

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 95 / 103
PARALLEL MANIPULATORS –
S INGULARITY IN FORCE DOMAIN
rank([gF ]) = 2 → Force ellipsoid shrinks to an ellipse and the Stewart platform
manipulator cannot apply a force normal to the plane of the ellipse.
rank([gF ]) = 1 or 0, the Stewart platform cannot apply any force in a plane or
cannot apply any external force, respectively.
Example:
Stewart-Gough platform with fixed base and moving platform as regular
hexagons of same size.
Consider the configuration of all legs parallel to the vertical.
[ H ] matrix is given by
 
0 0 0 0 0 0
 0 0 0 0 0 0 
 
 1 1 1 1 1 1 
[H] = 



 b1y b2y b3y b4y b5y b6y 
 −b1x −b2x −b3x −b4x −b5x −b6x 
0 0 0 0 0 0
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 96 / 103
PARALLEL MANIPULATORS –
S INGULARITY IN FORCE DOMAIN

Three rows of ([ H ]) are zero and the Stewart-Gough platform is in a singular

configuration → [gF ] has rank 1.
Tip of the force vector F can only lie along a line and only the vertical external
force can be resisted.
The Stewart-Gough platform, in this configuration, has singularity along Fx and
Fy .
Similar analysis can be done for M → Any Mz cannot be resisted in this
configuration.
Above analysis used by Ranganath et al. 2004 to design sensitive 6 components
force-torque sensors.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 97 / 103
R ELATION BETWEEN GAIN AND
FORCE SINGULARITY

If det([ K ∗ ]) = 0 → Parallel manipulator gains one (or more) degree of

freedom instantaneously.
If det([ H ]) = 0 parallel manipulator cannot resist forces/moments in one (or
more) directions at that configuration.
Relationship between two?
Illustrated using simple planar 4-bar mechanism!

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 98 / 103
R ELATION BETWEEN GAIN
F
φ3
AND FORCE SINGULARITY O2
Link 2
O3

φ2
ŶL
d l3 ŶR
Link 1 Link 3
{L} l1
{R}

τ1 φ1
θ1 X̂L
l0
X̂R
OR
OL , O1 Fig a

Ŷ O3
T2
T1 γ
l2 l3

O2 α1
R1 α2 OR
X̂

R2 Fig b

F IGURE : Static force analysis in a four-bar mechanism

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 99 / 103
R ELATION BETWEEN GAIN
AND FORCE SINGULARITY

With θ1 locked → Point O2 is fixed.

For a given θ1 , l0 and l1 , the length d opposite to θ1 is known.
Draw the planar truss structure determined by Link 2, Link 3 and the now fixed
side O2 − OR . Angles α1 and α2 can be computed in terms of θ1 , ϕ1 and ϕ2 .
Consider a force F = (Fx , Fy )T , acting at an angle β at point O3 .
Denote axial forces along the links O2 − O3 and O3 − OR of the planar truss by
T1 and T2 , respectively.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 100 / 103
R ELATION BETWEEN GAIN
AND FORCE SINGULARITY

By using a free-body diagram

( ) [ ]( )
Fx cos α1 − cos α2 T1
=
Fy sin α1 sin α2 T2

T1 and T2 are
( ) [ ]( )
T1 1 sin α2 cos α2 Fx
=
T2 sin(α1 + α2 ) − sin α1 cos α1 Fy

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 101 / 103
R ELATION BETWEEN GAIN
AND FORCE SINGULARITY
From T1 obtain reactions R1 and R2 at the joint O2
( ) [ ]( )
R1 1 cos α1 sin α2 cos α1 cos α2 Fx
=
R2 sin(α1 + α2 ) sin α1 sin α2 sin α1 cos α2 Fy

Torque required in joint 1, τ1 , to keep the four-bar mechanism in equilibrium

τ1 = R1 l1 s1 − R2 l1 c1
det([ H ]) = 0 if sin(α1 + α2 ) = 0 or γ is π radians → ϕ3 2π → Link 2 and Link
3 are aligned.
Gain singularity condition same as force singularity condition!
Gain singularity → Instantaneous velocity perpendicular to line along Link 2
and Link 3 (singular direction).
Force singularity → Any force along singular direction gives rise to infinite R1
and R2 and infinite τ1 is required to maintain static equilibrium.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 102 / 103
S UMMARY

Statics: All (actuated) joint locked and manipulator becomes a structure!

Find reaction forces and moments at joints due to externally applied force and
moment — Use of free body diagram.
Propagation of force/moments in serial manipulators similar to propagation of
velocities.
Serial manipulator: joint torques are related to external force/moment by
transpose of manipulator Jacobian.
Parallel manipulator — Force transformation matrix.
Loss of rank of Jacobian or force transformation matrix – Singularity.
Force/moment applied along singular direction cannot be resisted.
Gain singularity in parallel manipulator identical to loss of rank of force
transformation matrix.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
A SHITAVA G HOSAL (IIS C ) ROBOTICS : BASICS AND A DVANCED C ONCEPTS NPTEL, 2020 103 / 103
Suggested Additional Reading – Week - 6
• S. Bandyopadhyay and A. Ghosal 2004, “Analysis of configuration space singularities of closed
loop mechanisms and parallel manipulators”, Mechanism and Machine Theory, Vol. 39,pp.
519-544.

• S. Bandyopadhyay and A. Ghosal 2004b, “Analytical determination of principal twists in

serial, parallel and hybrid manipulators using dual vectors and matrices”, Mechanism and
Machine Theory, Vol. 39, pp. 1289-1305.

• D. Basu and A. Ghosal 1997 “Singularity analysis of platform-type multi-loop spatial mech-
anism”, Mechanism and Machine Theory, Vol. 32, pp. 375-389.

• A. Ghosal and B. Ravani 1998, “A dual ellipse is a cylindroid”, Proc. of ASME DETC’98,
Atlanta, USA, September 1998.

• A. Ghosal and B. Ravani 2001, “Differential geometric analysis of singularities of point trajec-
tories of serial and parallel manipulators”, Trans. of ASME, Journal of Mechanical Design,
Vol. 123, pp. 80-89.

• C. Gosselin and J. Angeles 1990, “Singularity analysis of closed loop kinematic chains”, IEEE
Trans. on Robotics and Automation, Vol. 6, pp. 281-290.

• K. H. Hunt 1986, “Special configurations of robot arms via screw theory, Part 1. The Jacobian
and its matrix cofactors”, Robotica, Vol. 4, pp. 171-179.

• K. H. Hunt, A. E. Samuel and P. R. McAree 1991, “Special configurations of multi-finger

multi-freedom gripper - A kinematic study”, The International Journal of Robotics Research,
Vol. 10, pp. 123-134.

• J.-P. Merlet 1991, “Singularity configurations of parallel manipulators and Grassman geom-
etry”, The International Journal of Robotics Research, Vol. 10, No. 2, pp. 123-134.

• F. C. Park and J. W. Kim 1999, “Singularity analysis of closed loop kinematic chains”, Trans.
of ASME, Journal of Mechanical Engineering Design, Vol. 121, No. 1, pp. 32-38.

• J. K. Salisbury 1982, Kinematics and force analysis of articulated hands, Ph. D. Thesis,
Stanford University.

• D. Zlatanov, R. G. Fenton, and B. Benhabib 1995, “A unifying framework for classification

and interpretation of mechanism singularities”, Trans. of ASME, Journal of Mechanical
Engineering Design, Vol. 117, No. 4, pp. 566-572.