Suggested solution INF3480 exam 2014

June 13, 2014

Exercise 1 (20 %)
a) The algorithm has the following steps.
(a) Initialize random population
(b) Evaluate individuals
(c) Check if termination criterion is reached
(d) Create new population from good individuals
b) Serial manipulators consist of one open chain, while parallel manipulators have a closed loop with
ground.

c) Any of the following will give a full score

To solve differential equations.
To analyze the stability of the system
To find the steady-state error using the final value theorem
To analyze the frequency response of the system (e.g. bode plot)
d) Typically mobile robots need exteroceptive sensors to sense the environment, while with manipu-
lators it is often sufficient to use only proprioceptive sensors.

Exercise 2 (45 %)
a) It is a PRP robot.
b) The DH parameters are

ai di i i
1 Lof f L1 + L2 + d1 0 0
2 0 0 90 2
3 0 L3 + L4 + d3 0 0

1
 In the rest of the exercise we will use

d1 = L1 + L2 + d1 (1)
d3 = L3 + L4 + d3 (2)

c) The transformations between the joints based on the DH parameters are (see (3.10) on page 77 in
the course textbook)

1 0 0 Lof f c2 0 s2 0
0 1 0 0 s2 0 c2 0
T10 = 1


T2 =
(3)
0 0 1 d1 0 1 0 0
0 0 0 1 0 0 0 1

1 0 0 0
2
0 1 0 0
T3 =
(4)
0 0 1 d3
0 0 0 1

Then we multiply these matrices together

c2 0 s2 Lof f
s2 0 c2 0
T10 T21 = T20 =


(5)
0 1 0 d1
0 0 0 1
s2 d3 + Lof f

c2 0 s2
0 2 0
s2 0 c2 c2 d3
T2 T3 = T3 = (6)
0 1 0 d1
0 0 0 1

d) The equations for the jacobian is given on page 133 in the textbook. It is a PRP manipulator,
therefore the jacobian is:
 
z0 z1 (o3 o1 ) z2
J= (7)
0 z1 0

2
 The calculating the elements in the jacobian.

s2 d3 + Lof f

Lof f
o3 o1 = c2 d3 0 (8)

d1 d1
s2 d3

= c2 d3 (9)
0

~i ~j ~k 
 

 
z1 (o3 o1 ) =  0 0 1  (10)
 s d c d 0 
2 3 2 3


c2 d3
= s2 d3 (11)
0

This yields the jacobian

c2 d3

0 s2

0 s2 d3 c2

1 0 0
J = (12)

0 0 0

0 0 0
0 1 0

e) The inverse kinematics can be solved analytically and geometrically. Solving analytically we start
with the tree equations

x = s2 d3 + Lof f (13)
y =c2 d 3 (14)
z =d1 (15)

It is already solved for d1 . Written with d1 yields

d1 = z L1 L2 (16)

Now aquaring the to remaining equations yields

 2
2
(x Lof f ) =s22 d3 (17)
 2
y 2 =c22 d3 (18)

Taking the sum of the two equations yields

 2
2
(x Lof f ) + y 2 = d3 (19)

The solving for d3

q
2
d3 = (x Lof f ) + y 2 L3 L4 (20)

Now only 2 is remaining. Rearranging the equations for x and y yields

x Lof f y
s2 = c2 = (21)
d3
d3

3
This can be solved using the atan2 function

3 = atan2(y, x + Lof f ) (22)

Since 2 does not depend on d3 an alternative solution for d3 can be found

y =c2 d3 (23)
y
d3 = (24)
c2

Inserting for d3 yields

y
d3 = L3 L4 (25)
c2

4
Exercise 3 (15 %)
a) Position of the mass is (can be found directly or by simplifying the forward kinematics in exercise
2)

L2 sin 2
p = L2 cos 2 (26)
d1 + L1

The finding the derivative (alternatively by simplifying the jacobian in exercise 2)

L2 c2 2

v = L2 s2 2 (27)
d1

The kinetic energy is

1
K= mv 2 (28)
2
1  
= m d21 + L22 22 (29)
2
Potential energy is (gravity in y-direction)

P = mgh (30)
= mgL2 cos 2 (31)

Combining these two to the Lagrangian yields

L=KP (32)
1  
= m d21 + L22 22 mgL2 cos 2 (33)
2

b) Using (7.5) on page 241 in the textbook, and calculating for the individual parts

L L
= md1 = mL22 2 (34)
d1 2
d L d L
= md1 = mL22 2 (35)
dt d1 dt 2
L L
=0 = mgL2 sin 2 (36)
d1 2
(37)

This yields the two dynamic equations

md1 = f1 (38)
mL22 2 mgL2 sin 2 = 2 (39)

In matrix form
   
m 0 0
q + = (40)
0 mL22 mgL2 sin 2

Exercise 4 (20 %)
a) Proportional controller (P-controller)

5
 b) The system without and with controller
Js2 = D K Js2 = D K + C(d ) (41)
The rearranging the equations yield
Js2 + K + D = 0 Js2 + (K + C) + D = Cd (42)
Then tranforming the equations into the time domain yields
d2 d2
J + K + D = 0 J + (K + C) + D = Cd (43)
dt2 dt2
c) Rearranging the equation (system with controller) from b) yields
Cd D
= (44)
Js2 + K + C
Inserting into the final value theorem
lim (t) = lim s(s) (45)
t s0

C
1s
T
1s
Dc
= lim
s (46)
s0 Js2 + K + C
CT Dc
= lim 0 (47)
s0 2
Js

* +K +C
C 1
= T Dc (48)
K +C K +C
d) Integral controller (I-controller / PI-controller)

1
CI s D

1 1 1
d
C
J s s

Exercise 5 (X %) (Master only)

a) Using (4.105) on page 149 in the textbook
= JT F (49)

2
3
0 0 1 0 0 0
1
1
= 2 3 12 0 0 0 1 (50)
0
21 1
2 1 0 0 0
0
0

1 3
= 3 2 (51)
3
2 31

6
b) The compliance frame is placed in the center of joint 2 with the x-axis pointing towards the tool:

First we define the natural constraints. The robot cannot move in the x-direction as it would
collide with the surface. There are no forces that can interact in the y- and z-direction. The robot
cannot rotate around the x and y axis, thus the velocities are zero. The is no external force when
rotating around the z-axis. This is summarized in the table below.
For the artificial constraints the force in the x-direction should be constant. The robot should
move to paint the entire surface, which means that the robot must have a variable velocity in vx
and z . The rest of the artificial constraints should be zero. The reciprocity condition is met with
these constraints.

Natural Artificial
vx = 0 fx = fc
fy = 0 vx = 0
fz = 0 vx = variable
x = 0 x = 0
y = 0 y = 0
z = 0 z = variable

c) Using the two equations provided in the exercise one get

G + Fe = M ax G + af
Mx (52)
= M ax + af Fe
Mx (53)
1
x
= ax + M (af Fe ) (54)
(55)

d) Using ax and af as they are defined in the textbook

d Md1 (Bd x
ax = x + Kd x
+ Fe ) (56)
af = Fe (57)

7
 Then insert this into the equation in c)

= ax + M 1 (af Fe )
x (58)
x
=x d M 1 (Bd x
d + Fe ) + M 1 (Fe Fe )
+ Kd x (59)
1
x
x =d

Md (Bd x
+ Kd x + Fe ) (60)

Md x
= (Bd x
+ Kd x
+ Fe ) (61)
+ Bd x
Md x + Kd x
= Fe (62)

We have now shown that the chosen ax and af gives the desired system.

e) Using the equation for the controller

u = M ax Gx + af (63)

= Mx d
Md1 (Bd x
+ Fe ) Gx + Fe
+ Kd x (64)

Then inserting the numbers

u = 2 0 1(0.5 0 + 0.25 0.1 + 2) 20 + 2 (65)

= (0.25 0.1 + 2) 20 + 2 (66)
= (0.25 0.1 + 2) 20 + 2 (67)
= 20.025 (68)