Velocity Analysis (Jacobian)

Velocity Analysis
• Needed to relate the joint velocities to the tool(end effector)velocities

1  X=FK(θ) x 
   y
 2  
 3  z 
 =  X = 
 4   
 5   
  θ =IK(X)  
 6   

Joint Space Task Space

Location of the tool can be specified using a joint space or a

Cartesian space description
 Velocity relations
• Relation between joint velocity and Cartesian velocity/end effector
velocity
• Jacobian matrix J(θ) 1  X = J ( )  x 
   y 
 2   
3   z 
   
 4   x 
   y 
  5
 = J −1
( ) X  
6   z 
Joint Space Task Space
 Jacobian
• Suppose a position and orientation vector of a manipulator is a
function of 6 joint variables: (from forward kinematics)
X = h(q)
x   q1   h1 ( q1 , q2 ,  , q6 ) 
 y q  h (q , q , , q )
   2  2 1 2 6 

z   q3   h3 ( q1 , q2 ,  , q6 ) 
X =   = h(   ) 6=  
 
1
q
 4 h
 4 ( q1 , q 2 ,  , q )
6 

   q5   h5 ( q1 , q2 ,  , q6 ) 
     
 
  
 q6 
 
 h6 ( q1 , q2 ,  , q6 ) 
 61
 Jacobean Matrix
Forward kinematics
X 61 = h(qn1 )

 d dh(q) dq dh(q)
X 61 = h(qn1 ) = = q
dt dq dt dq
 x 
 y   q1 
  q 
 z   dh(q)   2  X 61 = J 6n q n1
 =  dq    
 x    6n
 y    J=
dh(q)
 
 z 
q n  n1 dq
 Cont.
 x 
 y 
   q1  Jacobian is a function of q,
 z   dh(q)  q 2  it is not a constant!
 =   

 x   dq  6n   
 y   
 q n  n1
 
 z   h1 h1 h1 
 q 
q2 qn 
 1 
 dh(q)   h2 h2

h2 
J =   =  q1 q2 qn 
 dq  6n      
 h h6 h6 
 6  
 q1 q2 qn  6n
 Cont.
 x 
 y  Linear velocity Angular velocity
   q1 
 z  V   x    x =   q 
q n1 =  2 

X = =   
 x   V =  y  
 =  y =   
 y   
 z   z =  
    q n  n1
 z 

The Jacobian Equation

X = J q
6n n1
 Example
• 2-DOF planar robot arm (x , y)

– Given l1, l2 , Find: Jacobian

2
 x  l1 cos1 + l2 cos(1 +  2 )  h1 (1 , 2 )  l2
 y  =  l sin  + l sin( +  )  = h ( , )
  1 1 2 1 2   2 1 2 
  x  1 
Y =   = J  1 l1
 2 
y  

 h1 h1 
   2  − l1 sin 1 − l2 sin(1 +  2 ) − l2 sin(1 +  2 )
J = 1 =
 h2 h2   l1 cos 1 + l2 cos(1 +  2 ) l2 cos(1 +  2 ) 
 1  2 
 Singularities
• The inverse of the Jacobian matrix cannot be calculated when
det [J(θ)] = 0

• Singular points are such values of θ that cause the determinant of

the Jacobian to be zero
 • Find the singularity configuration of the 2-DOF
planar robot arm
determinant(J)=0 Not full rank

 x 
X =   = J  1 

 y   2  V (x , y)

− l1 sin 1 − l2 sin(1 +  2 ) − l2 sin(1 +  2 )

J = 
 1
l cos  1 + l 2 cos(1 +  2 ) l 2 cos(1 +  2 
) l2
Y
2 =0
Det(J)=0 Loss of 1 or more DOF
1 l1
2 = 0 x
 Singularity
at Singularities
✓ Singularities represent configurations from which certain directions
of motion may be unattainable
✓ The manipulator end effector cannot move in certain directions
✓ Singularities usually (but not always) correspond to points on the
boundary of the manipulator workspace(that is, to points of
maximum reach of the manipulator)
✓ Bounded End-Effector velocities may correspond to unbounded
joint velocities
✓ Bounded joint torques may correspond to unbounded End-Effector
forces and torques
 Jacobian Matrix
• Pseudoinverse
– Let A be an mxn matrix, and let A+ be the Pseudoinverse
of A. If A is of full rank, then A+ can be computed as:

 AT [ AAT ]−1 mn

+  −1
A = A m=n
[ AT A]−1 AT mn

 Cont.
– Example: Find X s.t.

1 0 2 3
1 − 1 0 x = − 2
   
1 1  −1 1 4 
+ T −1   5 1  1 
A = A [ AA ] = 0 − 1 
T
 = 1 − 5
1 2  9 
2 0  4 − 2

Matlab Command: pinv(A) to calculate A+

− 5
+ 1 
x = A b =  13 
9
 16 
 Inverse Jacobian
• Inverse Jacobian
 J11 J12  J16   q 
1

J  
q 
J 22  J 26  q 
2

X = Jq =  21
q = J X
3
 
−1       q  4

  q 
5

 J 61 J 62  J 66  q 
6

• Singularity
– rank(J)<n : Jacobian Matrix is less than full rank
– Jacobian is non-invertable
– Boundary Singularities: occur when the tool tip is on the surface of the
work envelop
– Interior Singularities: occur inside the work envelope when two or more
of the axes of the robot form a straight line, i.e., collinear
 Jacobian Matrix
 ax   bx 
• If A =  a y  , B = by 
 az   bz 

• Then the cross product

i j k  a y bz − az by 
 
A  B = ax ay az =  −(axbz − az bx ) 
bx by bz  axby − a y bx 
 Remember DH parameter
ci -c i si s i si a i ci 
 s ci c i -s i ci a isi 
Ai =  i
 0 s i c i di 
 
 0 0 0 1 
• The transformation matrix T

Ti = A1 A2 ..... Ai
0
 Jacobian Matrix
J = J1 J 2 .... J n 
 Example (Jacobian Matrix)
2-DOF planar robot arm ci -c isi s isi a i ci 
Given l1, l2 , Find: Jacobian  s c c -s c a s 
Ai =  i i i i i i i

• Here, n=2,
0 s i c i di 
 
0 0 0 1 
DH parameter (x , y)

2
l2

1 l1
 0 
Z 0 = Z1 = 0 
1 

0  a1 cos 1   a1 cos 1 + a2 cos(1 +  2 ) 

O0 = 0  , O1 =  a1 sin 1  , O2 =  a1 sin 1 + a2 sin(1 +  2 ) 
0   0   0  19
 Jacobian Matrix
2-DOF planar robot arm
Given l1, l2 , Find: Jacobian
• Here, n=2

(x , y)
 z0  (o2 − o0 )   z1  (o2 − o1 ) 
J1 =   , J2 =  
 z 0   z1  2
l2

1 l1
 Jacobian Matrix
 z0  (o2 − o0 ) 
J1 =  
 z 0 
 0   a1 cos 1 + a2 cos(1 +  2 ) 
Z 0  (o2 − o0 ) =  0    a1 sin 1 + a2 sin(1 +  2 ) 
1   0 
 i j k
=  0 0 1 
 a1 cos 1 + a2 cos(1 +  2 ) a1 sin 1 + a2 sin(1 +  2 ) 0 
 − a1 sin 1 − a2 sin(1 +  2 ) 
=  a1 cos 1 + a2 cos(1 +  2 ) 
 0 
 Jacobian Matrix
 z1  (o2 − o1 ) 
J2 =  
 z1 
0   a2 cos(1 +  2 ) 
Z1  (o2 − o1 ) = 0    a2 sin(1 +  2 ) 
1   0 
 i j k
=  0 0 1 
 a2 cos(1 +  2 ) a2 sin(1 +  2 ) 0 
 −a2 sin(1 +  2 ) 
=  a2 cos(1 +  2 ) 
 0 
 Jacobian Matrix
 − a1 sin 1 − a2 sin(1 +  2 ) 
 a cos  + a cos( +  )   − a2 sin(1 +  2 ) 
 1 1 2 1 2   a cos( +  ) 
 0   2 1 2 

J1 =    0 
 0  J2 =  
 0 
 0 
   0 
 1   
 1 

The required Jacobian matrix J

J =  J1 J2 
 Inverse Velocity
– The relation between the joint and end-effector
velocities:
X = J (q )q

where j (m×n). If J is a square matrix (m=n), the joint

velocities: −1
q = J (q ) X
– If m<n, let Pseudoinverse J+ where
+ T −1
+
q = J (q ) X J = J [ JJ ]
T
 Acceleration
– The relation between the joint and end-effector
velocities:
X = J (q )q
– Differentiating this equation yields an expression for the
acceleration d
X = J (q)q + [ J (q )]q
dt
Given X of the end-effector acceleration, the joint acceleration q
d
J (q)q = X − [ J (q )]q
dt
−1 d
q = J (q)[ X − J (q )]q ]
dt