Robotics 1

July 8, 2024

Exercise 1 Franka Research 3

Release Version: 1.1 (August 2022)
Based on the data sheet of the 7R robot shown in Fig. 1, assign the link frames and fill
Document in the
number: asso-
120020
ciated table of parameters according to the Denavit-Hartenberg (D-H) notation (use the attached
sheet). The joint axes are labeled from A1 to A7. Frame 0 and frame 7 are already displayed;
the green symbols ⊗ and denote here an axis
DIMENSIONS going in or, respectively, coming out the sheet, so
& WORKSPACE
as to complete a right-handed frame. The assignment of the D-H frames should be such that all
constant parameters are non-negative. Provide the numerical values of all parameters, including
those of the joint variables θi ∈ (−π, π], for i = 1, . . . , 7, in the configuration shown.

82 384 88

A5 A6

82 107
A4 A7 Y7
X
7

A3
316
Z7

A2

ZZ0
333
A1

Y0 X0

Axes names with joint lengths [mm]

Figure 1: From the data sheet of a 7R robot. Lengths are in [mm].

Exercise 2
In the orientation specified by the rotation matrix
855 855
 √ √ 
√2/2 0 √2/2
R =  2/2 0 − 2/2  ,
0 1 0
ISO CUBE

the end effector of a robot has the angular velocity

 
1188 1
ω =  1  [rad/s].
2 R 805
ge // to base
reach with flan

Represent the orientation R with RPY-type φ = (α, β, 333

angles
55γ)ofaround the sequence of fixed axes
ge
flan
R 8x. reach
ISO CUBE
YZX and provide the value φ̇ that produces the given angular
ma velocity.
362 365
1

280 280

Workspace | side view [mm] Workspace | top view [mm]

Exercise 3
Two equal 2R planar robots with unit length links share a collaborative task. With reference to
Fig. 2, the base of robot A is placed at the origin of the world frame while robot B is mounted
head down on the line Γ with a base that can slide on it. The line Γ is tilted by an angle γ = 135◦
with respect to the xw axis and intersects this axis at a distance ∆ = 4 m from the origin of the
world frame. When robot A is in the configuration q A = (π/4, −π/3) [rad], determine the position
of the base of robot B on Γ and its configuration q B such that the end effectors of the two robots
are in the same position, aligned and facing each other. Is the solution found unique?

yw

robot B

robot A
𝛾

xw
∆

Figure 2: Set-up for the collaborative task.

Exercise 4
Consider a rest-to-rest trajectory planning problem for a RP planar robot. The robot should
move its end effector along a linear path between the two Cartesian positions pi = (0.6, −0.3) and
pf = (−0.3, 0.6) [m], using a trapezoidal speed profile. The velocities of the two joints are bounded
by |q̇1 | ≤ 2 rad/s and |q̇2 | ≤ 1 m/s, while the acceleration along the path is bounded in norm as
kp̈k ≤ A = 0.5 m/s2 . What is the minimum feasible motion time T for this task? Provide also the
corresponding value of the joint velocity q̇ at the midpoint of the path.

[210 minutes (3,5 hours); open books]

2
 Franka Research 3
Solution Release Version: 1.1 (August 2022)
July 8, 2024 Document number: 120020

Exercise 1
DIMENSIONS
The 7R robot in Fig. 1 is a Franka Research&3.WORKSPACE
A correct assignment of D-H frames satisfying
the requests is shown in Fig. 3, while Tab. 1 contains the corresponding (non-negative) constant
parameters, as well as the values of the joint variables θ in the configuration shown. The axes z1 ,
z3 and z5 are coming out the sheet (denoted with ).

82 384 88

X4
A5 A6
Z4 Z5 Y5 X6
⨀
Y3 Z6
82 107
⨀ X3 X5
Z3 A4 A7 Y7
X
7

A3
316 Z2
Z7
Y1
X2 X1
⨀ A2
Z1

Z Z0
333
A1

Y0 X0

Axes names with joint lengths [mm]

Figure 3: D-H frames for the 7R Franka Research 3 robot.

i αi ai di θi

1 π/2 0 333 0

855 855
2 π/2 0 0 π

3 π/2 82 316 π

4 π/2 82 0 π/2
ISO CUBE

5 π/2 0 384 π

1188 6 π/2 88 0 π/2

7 0 0 107 0
R 805
ge // to base
reach with flan

333
Table 1: D-H parameters tor the frame assignment in Fig. ge
5 3 an(units in [rad] or [mm]).
85 of fl R ch
rea
ISO CUBE
x.
ma

362 365

3
280 280

Workspace | side view [mm] Workspace | top view [mm]

Exercise 2
The rotation matrix associated to the RPY-type angles φ = (α, β, γ) around the sequence of fixed
axes YZX is given by
 
cα cβ −sβ sα cβ
RYZX = RX (γ)RZ (β)RY (α) =  cα sβ cγ + sα sγ cβ cγ sα sβ cγ − cα sγ 
cα sβ sγ − sα cγ cβ sγ sα sβ sγ + cα cγ

The inverse relations from an orientation matrix R = {Rij } to (α, β, γ) are given by
     
R13 R11 R32 R22
q
2
β = atan2 −R12 , ± R11 + R13 2 α = atan2 , γ = atan2 , ,
cβ cβ cβ cβ
p
out of the representation singularity cβ = R112 + R2 = 0.
13

For the given rotation matrix R, this gives the two regular solutions
       
α1 π/4 α2 −3π/4
φ1 =  β1  =  0  φ2 =  β2  =  π .
γ1 π/2 γ2 −π/2

The contributions of the three time derivatives α̇, β̇ and γ̇ to ω when the orientation is φ is
computed as1
     
1 0 0
ω = ω γ̇ + ω β̇ + ω α̇ =  0  γ̇ + RX (γ)  0  β̇ + RX (γ)RZ (β)  1  α̇
0 1 0

and thus ω = T (β, γ)φ̇ with

 
−sβ 0 1
T (β, γ) =  cβ cγ −sγ 0 .
cβ sγ cγ 0

Note that det T (β, γ) = cβ vanishes exactly at the singularity of the YZX RPY-type representation.
Evaluating T for the two solution triples φ1 and φ2 gives
   
0 0 1 0 0 1
T 1 = T (β1 , γ1 ) =  0 −1 0  T 2 = T (β2 , γ2 ) =  0 1 0  .
1 0 0 1 0 0

Accordingly, we have the two solutions

  
2 2
φ̇1 = T −1
1 ω =
 −1  φ̇2 = T −1
2 ω =
 1 .
1 1

1 An alternative but more lengthy way would be to use the relationship Ṙ(φ)RT (φ) = S(ω), and then extracting

from the off-diagonal elements of the skew-symmetric matrix S the components ωx , ωy and ωz . Also, the elements
of matrix Ṙ have a linear dependence on the components α̇, β̇ and γ̇ of φ̇.

4
Exercise 3
The problem involves the use of a suitable homogeneous transformation between the base frames
of the two robots and can be solved in different ways. The following is a simple one.
Through its direct kinematics, the end-effector position of robot A for q A = (π/4, −π/3) [rad] is
   
w cos qA1 + cos(qA1 + qA2 ) 1.6730
pA = f A (q A ) = = [m].
sin qA1 + sin(qA1 + qA2 ) 0.4483

as expressed in the world frame, which coincides with the base frame of robot A. Robot B should
place its end effector in this same position, facing the end effector of robot A and with an orientation
that is aligned with the second link of this robot. Thus, it is convenient to extend the robot A by
adding to its second link also the length of the second link of robot B (all links have unit length),
namely with the modified direct kinematics
   
w cos qA1 + 2 cos(qA1 + qA2 ) 2.6390
pE = = [m].
sin qA1 + 2 sin(qA1 + qA2 ) 0.1895

The position pE is shown in Fig. 4. This point should be the target for the tip of the first link
of robot B, without further conditions on the orientation part of the collaborative task (already
satisfied by the ‘trick’ of extending the second link of robot A). Accordingly, the robot B mounted
on a sliding base and taken up to the tip of the first link can be seen as an equivalent PR robot
with a fixed base placed at the intersection between the line Γ and the world axis xw .

yw

qB1
qA2 xb
pA = pB 𝛾
xB
pE
qA1

qB2 xw
yB

Figure 4: Graphical illustration of the two solutions for the collaborative task.

Place then the base frame of robot B as in Fig. 4. The homogeneous transformation2 between the
2 Since the problem is planar, we will use here a 3×3 homogeneous matrix, with a 2×2 rotation matrix R ∈ SO(2)

and a position vector p ∈ R2 .

5
base frames A (≡ w) and B is then
   √ √ 
cos γ − sin γ ∆ − 2/2 − 2/2 4
w A   √ √
T B = T B =  sin γ cos γ 0 = 2/2 − 2/2 0 .
 
0 0 1 0 0 1

As a result, we compute
 w

B B w w −1 pE
pE,hom = T w pE,hom = TB
1
 √ √    
− 2/2 2/2 2.8284 2.6390 1.0964 B
 √ √
 
pE
=  − 2/2 − 2/2 2.8284   0.1895  =  0.8284  = .
   
1
0 0 1 1 1

The direct kinematics of the equivalent PR robot is simply

 
B xb + cos qB1
pBeq = f PR (xb , qB1 ) = .
sin qB1

The inverse kinematics is given by

q
xb = B pBeq,x ± 1 − B pBeq,x
B
qB1 = atan2 pBeq,y ,B pBeq,x − xb .

The joint angle qB2 of robot B is found by setting the difference between the absolute orientations
of the two end effectors so that they face each other; i.e., the difference should be equal to π:

(γ + qB1 + qB2 ) − (qA1 + qA2 ) = π ⇒ qB2 = qA1 + qA2 + π − (γ + qB1 ) .

Setting now B pBeq = B pE = (1.0964, 0.8284) [m], two solutions are found, as sketched graphically
in Fig. 4; one solution is closer to the base frame of robot B
 
1.0343
xb = 0.6037 m qB = [rad],
−0.5107

while the other is further away

 
2.3186
xb = 1.8647 m qB = [rad].
−1.7950

Exercise 4
The RP planar robot3 and the desired motion task are shown in Fig. 5. The linear path has length
L = pf − pi = 1.2728 m and is traced by

pf − pi
p(s) = pi + s s ∈ [0, L],
L
where s is the path parameter (here, the arc length). The timing law s(t), for t ∈ [0, T ], should
have a rest-to-rest (symmetric) trapezoidal profile for the speed ṡ, which is fully described by the
3 This is by default the most common structure of a PR planar robot. Moreover, we shall assume that q2 > 0.

6
 pf

q2

q1
x

pi

Figure 5: Trajectory planning task for a RP robot.

cruising speed V along the path and by the acceleration A in the first phase (with a rise time
Tr = V /A). The Cartesian velocity and acceleration of the robot end effector are, respectively,
pf − pi pf − pi
ṗ = ṡ p̈ = s̈.
L L
While we have for the acceleration norm kp̈k = |s̈| ≤ A = 0.5 m/s2 , there is instead no explicit
bound specified in the Cartesian space for the velocity norm kṗk = |ṡ| ≤ V . This should be derived
from the available velocity limits in the joint space.
Note first that, using the inverse kinematics of the RP robot
q
q1 = atan2 {py , px } q2 = kpk = p2x + p2y ,

we obtain from the initial and final Cartesian points pi and pf

−0.4636
   
2.0344
qi = [rad, m] qf = [rad, m].
0.6708 0.6708
The revolute (first) joint has to travel by ∆q1 = qf,1 − qi,1 = 2.4981 rad. Therefore, its motion
time is lower bounded by |∆q1 |/V1 = 1.2490 s (assuming an infinite joint acceleration). Moreover,
the joint value at the midpoint is qm,1 = (qi,1 + qf,1 ) /2 = 0.7854 rad. On the other hand, since
qf,2 = qi,2 = 0.6708, the prismatic (second) joint needs first to reduce its length in order to remain
on the linear Cartesian path, and then to reverse motion increasing the length back to the initial
value in a symmetric

way with respect to the path midpoint; the minimum extension will be at
pm = pi + pf /2 = (0.15, 0.15) m, corresponding to qm,2 = 0.2121 m. Therefore, the motion
time of the second joint is lower bounded by (|qm,2 − qi,2 | + |qf,2 − qm,2 |)/V2 = 0.9174 s (assuming
again an infinite joint acceleration).
The above analysis shows that the limiting velocity factor is due to the revolute joint. As a result,
we can take as upper bound for the Cartesian speed along the path the worst case situation, namely
when the distance to the path is minimum, i.e., at qm,2 = 0.2121 m, and evaluate
|ṡ| ≤ V = qm,2 · V1 = 0.4243 m/s.

7
 With L, V , and A, we compute the minimum feasible motion time along the linear path when
using a trapezoidal profile4 as

LA + V 2 L V
T = = + = 3.8485 s.
VA V A

To evaluate the joint velocity q̇ at the path midpoint (corresponding to q m = (0.7854, 0.2121) [rad,m]),
we need the task Jacobian for this robot:
√ !
−q2 sin q1 cos q1 −0.15
 
2/2
J (q) = ⇒ J m = J (q m ) = √ .
q2 cos q1 sin q1 0.15 2/2

Being ṗm = V (pf − pi )/L = (−0.3, 0.3) [m/s] (the speed at the path midpoint is certainly at the
cruise value), we have as expected
 
−1 2
q̇ m = J m ṗm = [rad, m].
0

Figure 6 shows the components of the planned trajectory in the Cartesian space and of the corre-
sponding trajectory in the joint space, together with their velocity and acceleration.

Cartesian position (x = blue, y = red)

0.5
[m]

-0.5

0 0.5 1 1.5 2 2.5 3 3.5

velocity

0.2
[m/s]

0
-0.2
-0.4
0 0.5 1 1.5 2 2.5 3 3.5
[s]
acceleration
0.5
[m/s 2]

-0.5
0 0.5 1 1.5 2 2.5 3 3.5

Figure 6: Left: The components of the minimum-time Cartesian trajectory using a trapezoidal
speed profile. Right: The components of the corresponding jont trajectory.

∗∗∗∗∗

4 Since L > V 2 /A, the existence of a motion phase at cruise speed V is guaranteed.