Robotics I

September 10, 2012

A 3R robot manipulator has the following Denavit-Hartenberg table:

i αi ai di θi
1 π/2 a1 > 0 0 θ1
2 0 a2 > 0 0 θ2
3 0 a3 > 0 0 θ3

Table 1: DH table of a 3R robot

1. Sketch the kinematic structure of the robot and place the D-H frames according to Table 1.

T
2. Draw the robot in the configuration θ = 0 π/4 −π/4 [rad].

Assume now the numerical data a1 = 0.2, a2 = 0.5, and a3 = 0.5 [m] and let the robot be in the
configuration specified at step 2.

T
3. Given a desired velocity v = 1 1 0.5 [m/s] for the robot end-effector (the origin O3
of frame 3), determine the instantaneous joint velocity vector θ̇ that realizes v.
4. With the solution θ̇ found at step 3, compute the associated angular velocity ω of the robot
end-effector frame.

5. Let the value ω found at step 4 be the desired angular velocity for the robot end-effector
frame. Characterize all instantaneous joint velocities θ̇ that realize ω at the given robot
configuration.
6. What is the structure of all feasible ω that can be realized by this robot in a generic config-
uration θ? What can we say about the differential mapping θ̇ → ω?

[120 minutes; open books]

1
 Solution
September 10, 2012

The robot has a kinematic structure similar to that of the first three joints of the KUKA KR5
robot (the industrial robot in our Robotics Laboratory). Figures 1 and 2 provide, respectively,
a sketch of the kinematic structure, with associated D-H frames, and the robot posture at the
specified θ.

y!3! x!3!

O!3!
z!3!
a!3!
z!0!
x!2!
a!2!
y!0!
a!1! z!2!
O!0!
x!1!
z!1!
x!0!

Figure 1: Kinematic structure and D-H frames

x!2! y!3!
z!0!
z!2! x!3!
O!3!
y!0!
z!3!

O!0!
x 0!= x!1!
z!1!
x!0!

T
Figure 2: The robot at the configuration θ = (0 π/4 −π/4)

For steps 3-6, we need to compute the robot Jacobian J (θ). For the linear part, J L (θ), we
may use either the vector product computations of the geometric Jacobian or simply differentiate
analytically the positional direct kinematics. From the product of the homogeneous matrices

2
associated to the D-H table 1, it follows
(a1 + a2 c2 + a3 c23 ) c1
 
     (a + a c + a c ) s
p 0

 1 2 2 3 23 1
phom = = 0A1 (θ1 ) 1A2 (θ2 ) 2A3 (θ3 ) = .

1 1  a2 s2 + a3 s23 
1
Therefore,
 
− (a1 + a2 c2 + a3 c23 ) s1 − (a2 s2 + a3 s23 ) c1 −a3 s23 c1
∂p
v = ṗ = θ̇ = J L (θ) θ̇, with J L (θ) =  (a1 + a2 c2 + a3 c23 ) c1 − (a2 s2 + a3 s23 ) s1 −a3 s23 s1  .
 
∂θ
0 a2 c2 + a3 c23 a3 c23

For the angular part, J A (θ), we have by definition (taking into account that velocity vectors are
expressed by default in the 0th frame)

J A (θ) = 0z 0 0z 1 0z 2 = 0z 0 0R1 (θ1 ) 1z 1 0R1 (θ1 ) 1R2 (θ2 ) 2z 2 ,

T
with iz i = 0 0 1 , for i = 0, 1, 2. As a result,
 
0 s1 s1
ω = J A (θ) θ̇, with J A (θ) =  0 −c1 −c1  . (1)
 

1 0 0

T
Evaluating the two Jacobians at the configuration θ = 0 π/4 −π/4 with the given
numerical data yields
   
0 −0.3536 0 0 0 0
J L =  1.0536 0 0 , J A =  0 −1 −1  . (2)
   

0 0.8536 0.5 1 0 0

T
Therefore, for v = 1 1 0.5 ,
   
0.9492 0
θ̇ = J −1
L v =
 −2.8284  [rad/s] ⇒ ω = J A θ̇ =  −3  [rad/s]. (3)
5.8284 0.9492

From the general structure of J A (θ) in (1) we see that this matrix is always singular, having
constant rank equal to 2. At a generic configuration (i.e., for a generic value of θ1 ), we characterize
the following subspaces of interest:
     
 0 s1   0 
R(J A (θ)) = span  0  ,  −c1  , N (J A (θ)) = span  1  .
1 0 −1
   

Therefore, all feasible ω will have the form

3
with α = θ̇1 ∈ R and β = θ̇1 + θ̇2 ∈ R. Conversely, given a generic θ̇ generating a ω, the same
value of end-effector angular velocity is obtained by adding a joint velocity vector θ̇ 0 ∈ N (J A (θ)),
or  
0  
θ̇ + γ θ̇ 0 = θ̇ + γ  1  ⇒ ω = J A (θ) θ̇ = J A (θ) θ̇ + γ θ̇ 0 .
−1
for any γ ∈ R.

T
Particularizing this general result to the specific configuration θ = 0 π/4 −π/4 , with
J A given in (2), all joint velocities that generate the same value ω as in (3) are given by
     
0.9492 0 0
θ̇ γ =  −2.8284  + γ  1  , for any γ ∈ R ⇒ ω = J A θ̇ γ =  −3  .
5.8284 −1 0.9492
∗
Note that the minimum norm joint velocity θ̇ realizing this value of ω is obtained by unconstrained
minimization of kθ̇ γ k2 with respect to γ. This yields
 
T 0.9492
θ̇ θ̇ 0 ∗
γ=− T = 4.3284 ⇒ θ̇ =  1.5  ,
θ̇ 0 θ̇ 0 1.5
∗
with kθ̇ k = 2.3240 —as opposed to kθ̇k = 6.5476 for the value θ̇ computed in (3). As could be
expected, the minimum norm solution balances the effort between the velocities of joints 2 and 3.

∗∗∗∗∗