Introduction to Robotics (CS223A) Homework #4 Solution

(Winter 2007/2008)

1. Consider the following RRRR manipulator (image courtesy J. J. Craig):

It has the following forward kinematics and rotational Jacobian:

2
c12 c34 s s
2 12 34
c12 s34 22 s12 c34 2
s
2 12
2c c s12 (s3 1) + c1
12 3
s12 c34 + 2
s12 s34+ 22 c12 c34 2
0 2 c12 s34 2c12 2s12 c3 + c12 (s3 1) + s1
4T =
2 2 2
2 s34 2 c34 s3 + 1
2

0 0 0 1

2 2
0 0 2 2
0
J = 0 0 0

0

2 2
1 1 2 2

(a) Find the basic Jacobian Jo in the {0} frame, for the position q = [0, 900 , 900 , 0]T .
(q is the vector of joint variables.)

h i
0 0 Pe 0 Pe 0 Pe 0 Pe
Jv = q1 q2 q3 q4

where 0 Pe is from the 4th column of 04 T . Thus:

2s12 c3 c12 (s3 1) s1 2s12 c3 c12 (s3 1) 2c12 s3 s12 c3 0
0
Jv = 2c12 c3 s12 (s3 1) + c1 2c12 c3 s12 (s3 1) 2s12 s3 + c12 c3 0
0 0 c3 0
 Plug in q = [0, 900 , 900 , 0]T , and join with 0 J (which was directly given to us) to get:

0 0 0 0
3 2 2 0



0J =
0 0 0 0

o
0 0 12 1

2

0 0 0 0
1 1 12 1
2

(b) A general force vector is applied to the origin of frame {4} and measured in
frame {4} to be [0, 6, 0, 7, 0, 8]T . For the position in (a), determine the joint
torques that statically balance it.
We are given a 6 1 force/moment vector Fapp which is exerted on the robot. If the
arm is statically balancing this, then we know that the robot must be exerting an equal
and opposite force/moment vector at the origin of frame {4} (we can thank Sir Isaac
Newton for that!).

So we know that in the coordinates of frame {4}, the vector 4 F4 = 4 Fapp and we want
to find the joint torques corresponding to 4 F4 .

Recall that = J T F . To multiply F and J, however, they must be in the same frame.
You can transform either the J from frame {0} to {4}, or transform F from frame {4}
to {0}. Both give the same answer.

4
F4 = 4 Fapp = [0, 6, 0, 7, 0, 8]T
" #
0R 0
0 4 4
F4 = 0R F4
0 4
0 T0
= J F4
The final answer is:

= [18.707, 12.707, 16.485, 8.0]T

(c) Consider the same configuration as above. A screw driver is gripped in the
end-effector so that its tip is along Z4 at a distance of 9 units of length from
the origin of frame {4}. What is the force and torque the screw driver tip
applies when the same joint torques that were determined in part (b) are
applied?
Lets look at the free-body diagram of the screw-driver, with the left-end being at origin
04 and the screw-driver tip on the right. NOTE: In this diagram, we consider 3x1 force
and moment vectors, so F represents the 3x1 linear force, NOT the combined 6x1
vector.
 We must first choose an origin for our computations, and then apply static equilibrium.
For this computation, the choice of origin is arbitrary! You should get the same answer
regardless. Two sensible options are either the origin of frame {4}, or the tip {S} of the
screw-driver. Lets use the origin of frame {4}. Also, for simplicity well express all our
vectors using the coordinates of frame {4}.
In static equilibrium, we know: F = 0 and N = 0. These give us:

F4 + (Fs ) = 0 Fs = F4
N4 + (Ns ) + Ps4 (Fs ) = 0 Ns = N4 +4 Ps (Fs )

The position 4 Ps is the position vector from origin {4} to the tip, so we know that: 4 Ps =
[0, 0, 9]T . Meanwhile, from part (b), we that: F4 = [0, 6, 0]T and N4 = [7, 0, 8]T . If
we first solve for Fs (using the upper equation), we can then use this value to solve for
Ns in the lower equation. THUS:

4 F = [0, 6, 0]T
s
4 N = [61, 0, 8]T
s

2. Consider the PRRP manipulator schematic shown below:

L2 L3
z0

x0

(a) Assuming no joint limits, sketch the workspace of this manipulator. Be sure
to include dimensions in your drawing. Assume L2 > L3 .
Since the prismatic joints have no limits, the workspace is an infinite cylinder along the
Z0 direction, whose cross-section is shown in the following figure.

L2 -L 3

L2 +L3

(b) Describe the (3D) dextrous workspace of this manipulator.

This manipulator can only point its end-effector downwards, so there are no points for
which it can achieve an arbitrary orientation. Even if you consider only the orientation
 with respect to the (X0 , Y0 ) plane (eg. the angle with the X0-axis) there are only two
joints to control the position in the plane, leaving no degrees of freedom for controlling
the orientation. Therefore, the dexterous workspace is null.
(c) With no joint limits, if we are considering only the position of the end effec-
tor, how many inverse kinematic solutions are there (in general)? Explain
briefly.
If we find a configuration of the joints in this manipulator that places the end-effector at
a given position, we can achieve the same position by shortening one prismatic joint and
extending the other by any value . This manipulator is redundant in the Z0 direction,
so an infinite number of inverse kinematic solutions exist.
(d) Imagine that we remove the first prismatic joint, so that the first revolute
joint now rotates around the base. Repeat part (c) for such an RRP manip-
ulator.
If we remove one of the prismatic joints, the manipulator is no longer redundant. For
any point (x, y, z), the extension of the prismatic joint is completely determined by z.
In the (X0 , Y0 ) plane, however, there are two values of the revolute joint angles that will
achieve a given (x, y), however: elbow up and elbow down. Therefore, there are two
inverse kinematic solutions for a given position.
(e) Imagine that we further modify the manipulator from part (d) by inserting
another revolute joint between the two existing revolute joints, whose axis
is oriented in the same direction as the other two. Repeat part (c) for such
an RRRP manipulator.
Compared to part (d), now the manipulator is redundant in the (X0 , Y0 ) plane. For a
given planar position (x, y), there are three revolute joints for only two position variables
(ie. for x and y), thus there are an infinite number of joint angles that will achieve it.
This means there are an infinite number of inverse kinematic solutions.

3. We wish to move a single joint from 0 to f , starting and ending at rest, in time
tf . The values of 0 and f are given, but we wish to calculate tf so that these
constraints hold: |(t)| < max and |(t)| < max for all t, where max and max are
given positive constants.

(a) Using a single cubic segment, give equations for the cubics coefficients ai in
terms of 0 , f and tf .
You can get this right out of the lecture notes or textbook. The long answer goes like
this: for the cubic polynomial (t) = a0 + a1 t + a2 t2 + a3 t3 , we have

(0) = a0 = 0
(tf ) = a0 + a1 tf + a2 t2f + a3 t3f = f
(0) = a1 = 0
(tf ) = a1 + 2a2 tf + 3a3 t2f = 0

Treating tf as a constant, the above is just a linear system of four equations and four
 unknowns (the ai s), and it can be solved with a little simple algebra to yield:

a0 = 0
a1 = 0
3( )
a2 = ft2 0
f
2(f 0 )
a3 = t3f

(b) Using the velocity constraint, |(t)| < max , derive a condition on tf in terms
of 0 , f , and max .
What we can say about |(t)| on the interval [0, tf ]? First of all, we know that (0) =
(tf ) = 0, so |(t)| must have its maximum value (in the interval [0, tf ]) at some extrema,
where (t) = 0. This is really just an extreme value problem from your first year calculus
class; we find the formula for (t) and set it equal to zero.
The polynomial (t) is given from part (a):
3(f 0 ) 2 2(f 0 ) 3
(t) = f + t t
t2f t3f

So, taking the first derivative yields

6(f 0 ) 6(f 0 ) 2
(t) = t t
t2f t3f

And the second derivative is

6(f 0 ) 12(f 0 )
(t) = t
t2f t3f

Setting (t) = 0 yields t = tf /2, which makes perfect sense: the velocity is quadratic,
and it has equal value at the endpoints of the interval, so its extreme value as at the
midpoint of the interval. So, we know that, on the interval [0, tf ],

|(t)| |(tf /2)|

So, in order to make sure that the condition on the maximum velocity is satisfied, we
need to make sure that

| (tf /2)| < max

6(f 0 ) 6(f 0 )
| 2 (tf /2) (tf /2)2 | < max
tf t3f
3(f 0 )
| | < max
2tf
3|f 0 |
< max
2tf

So, our condition is

3|f 0 |
tf > 2max
(c) Using the acceleration constraint, |(t)| < max , derive a condition on tf in
terms of 0 , f , and max .
This problem is completely analagous to part (b), except that the acceleration is linear,
so it will achieve its extreme value at one of the endpoints of the interval. If we plug in
t = 0 and t = tf into the acceleration equation, we get

6(f 0 )
(0) =
t2f
6(f 0 )
(tf ) =
t2f

So, we know that, on the interval [0, tf ],

6(f 0 ) 6|f 0 |
|(t)| | 2 |=
tf t2f

So, in order to make sure that the condition on the maximum acceleration is satisfied,
we need to make sure that
6|f 0 |
< max
t2f

So, our condition is

r
6|f 0 |
tf > max