Exam for DD2410 - Introduction to robotics 2021-10-28

Permitted aids: Paper, pencil, eraser

Notes:

Answer on the separate answer sheet. Write legibly, ambiguous answers

will be given 0 points.

Each question can be awarded a maximum of 4 pts. If a question can be

awarded partial credit, this will be explicitly described in the question.

Maximum possible score on the exam is 100 pts. Any bonus points will
be added to the exam score.

A passing grade requires a sum of exam score and bonus points of at

least 80 pts.

A sum of exam score and bonus points of at least 70 pts will be graded
as Fx, which means that the student is given the opportunity of
completing the grade to a passing grade after passing an oral exam.

Course responsible: Christian Smith ([email protected])

Question 1:

1 (21)
For each of the robot applications below, select the most suitable
actuator type. Choose from the provided list, (1 p for each correct
answer).

Possible actuators:
1) Pneumatic artificial muscle
2) Piezoelectric artificial muscle
3) Permanent magnet brushless motor
4) Hydraulic cylinder
5) Pneumatic cylinder
6 Electric stepper motor
7) Variable impedance antagonistic actuator

Applications:

a) A low-cost robot that uses a laser to engrave letters onto metal

plates.6

b) A construction site robot that can lift and hold heavy workpieces
forlonger periods of time.4

c) A high precision welding robot for assembling electric cars.3

d) A high-speed parallel pick-and-place robot for placing components

inthe right compartments of cases passing on a conveyor belt.3

Question 2:

How many degrees of freedom can the end effector of a robot

manipulator with the following configurations at most be moved in?
Answer with a number for each configuration.
(2p for each correct answer)

a) PPPPPPP (i.e. 7P) 3

b) RRRRRRR (i.e. 7R) 6

Question 3:

Which of the following sensor combinations (a-d) would be sufficient to

enable a hybrid force/position controller, where the robot end effector is

name: _______________________________________________________ number:______________________________

2 (21)
to move along a given trajectory on a known surface, and apply a given
normal force? Assume a 6 DoF robot with revolute joints, similar to a
Puma 560, with known kinematics and dynamics, where we can supply
arbitrary voltages to the motors. Assume no singularities in the
workspace.
(4p for correct list of combinations.)

a) Optical joint encoders, 3 DoF accelerometers at the end effector, and

aHall effect sensor in the tool tip.

b) Hall effect joint encoders, strain gauge torque sensors at all joints,
anda gyroscope in each link.

c) A 6 DoF optical force/torque sensor at the robot wrist, Hall effect

jointencoders, and ammeters on all motor power cables.

d) 3 DoF Accelerometers in all links, a Hall effect sensor at the tool

tip,and Optical joint encoders.

e) 6 DoF strain gauges in all links, Hall effect joint encoders, and
torquesensors on all motor drive shafts.

Question 4:

List the generally correct statements regarding homogeneous

transforms.
(4p for completely correct list. 2p if one correct statement is missing. 0p if
more than one correct statement is missing, or an incorrect statement is
included)

Assume:

pi : point p in frame i

: time derivative of point p in frame i

a
Tb : coordinate transform (homogeneous transformation) between
frames a and b

a) pa = aTbpb

name: _______________________________________________________ number:______________________________

3 (21)
b) pa = (bTa)-1pb

c) pa = (bTa)Tpb

d)

e)

f) pa = aTxxTbpb

g) pa = (bTxxTa)-1pb

h) pa = (xTbaTx)Tpb

Question 5:

Mark all statements regarding the inverse kinematics of

robotic manipulators as either True or False. (1p per correctly
classified statement.)

a) There exist no solutions to the inverse kinematics for

singularconfigurations of the robot.F

b) If we have the complete Denavit-Hartenberg parametrization of

themanipulator, we can always formulate the inverse kinematics in an
analytical closed form.F

c) For some robots with only prismatic joints, the inverse kinematics is
alinear function of the configuration.T

d) There can exist several different solutions for the inverse

kinematicsfor a specific end effector pose.T

Question 6:

What is the minimum number of contact points necessary to generate

a force closure grasp on the depicted 3D shape, assuming sufficient
friction? Answer with a number. (The correct answer gives 4 p)

name: _______________________________________________________ number:______________________________

4 (21)
3

Question 7:

Order the following operations with regard to computational cost, from

the one requiring the most effort to compute to the least. We are
considering a generic 6 DoF RRRRRR manipulator. (Complete list in
correct order gives 4 p) dacb

a) Calculate the inverse Jacobian given the DH table and current

jointvalues.

b) Calculate the coordinate transform from the base frame to the

endeffector frame, given the DH table and current joint values.

c) Calculate the torque load on the joints caused by an external load

onthe end effector, given the DH table, current joint values, and the
external load.

d) Calculate the inverse kinematics for a desired end effector pose,

giventhe DH table, current joint values, and the desired pose.

Question 8:

Qualify each of the following sensor types as either proprioceptive (P) or

exteroceptive (E).
(Each correctly marked alternative gives 1 p) all P

a) A joint torque sensor.

b) An ammeter measuring motor currents.

name: _______________________________________________________ number:______________________________

5 (21)
 c) An optical joint encoder.

d) A thermometer measuring temperatures in the reduction gears.

Question 9:

Given a 7 DoF (RRRRRRR) manipulator (as in the image below), which

of the following pairs of quantities should we expect to be able to control
independently? Assume that we have the appropriate sensors.
(4 pts. for correct list)

a) Joint torques and position of the end effector

b) Orientation and position of the end effector.

c) Interaction forces and velocity of the end effector.

d) Orientation and interaction forces of the end effector

e) Acceleration and velocity of the end effector

f) None of the above

Question 10:

For each statement on manipulator control, mark if it is true (T) or

false (F).
(1 p for each correctly classified statement)

name: _______________________________________________________ number:______________________________

6 (21)
 a) Computed Torque Control is very sensitive to the quality of the
dynamical models of the manipulator, but can give very good
performance at the dynamical limits of the system.T

b) Controlling each joint independently with a local PID controller is

very sensitive to the quality of the dynamical models of the
manipulator, but can give very good performance at the dynamical
limits of the system.F

c) Using a reasonably tuned PID feedback controller for the whole

arm, the I (integral) component will eventually compensate for the
effects of gravity within a local region.T

d) If all angles ɑi in the DH table are integer multiples of 90 o, the

motion of different joints will not interfere with each other, so that
we can control all joints independently without loss of dynamic
performance.F

Question 11:

Below you can find the pseudocode for RRT* from the original 2011
paper, with line numbers given to the left. To convert the RRT* to an RRT
algorithm, you have to remove a number of lines, and modify line 13
somewhat.

a) List all line numbers of lines to be removed: (3p for a

correctanswer) 7-12 14-16

b) What modifications are needed for line 13? (1p for a

correctanswer) ....Change “x_min” to “x_nearest"

name: _______________________________________________________ number:______________________________

7 (21)
Question 12:
Consider the problem of planning a path between the two stars shown in
the figure below. Black cells are non-traversable obstacles. All grid cells
are 1m wide squares.
(1 point for each correct answer) Q12: a)...17±1 m.b).No.....c) 5 + 6 2 ±1 m
d)..No..

a) Draw a path that might be produced by A* on the grid

graphwhere each grid cell is connected to 4 neighbors (north, east,
west and south). How long is it? Answer with a number.

b) Does the problem in A have a unique optimal solution?

Answeryes or no.

c) Draw a path that might be produced by A* on a grid graph

whereeach grid cell is connected to 8 neighbors (the 4 above as

name: _______________________________________________________ number:______________________________

8 (21)
 well as north-west, north-east, south-west and south-east.). How
long is it? (assuming diagonal edges are sqrt(2) m). Answer with a
number.

d) Does the problem above have a unique optimal solution?

Answeryes or no.

Question 13:

Consider once again the figure of Question 12. Q13: a).shorter....

b)..yes………c).....1 m… d).19 ±1 m.

a) Draw a path that might be produced by A* on a visibility graph

created from the obstacles. Is it shorter or longer than the
path in 12c?

b) Does the problem in 13 a have a unique optimal solution?

c) Imagine the robot is a 3m by 3m rectangular block. Draw the

configuration space obstacles in the figure. How narrow is the
most narrow passage in the configuration space (in m)?
Answer with a number.

d) Draw a path that might be produced by A* on a grid graph,

using the configuration space obstacles, where each grid cell
is connected to 4 neighbors (north, east, west and south).

name: _______________________________________________________ number:______________________________

9 (21)
 How long is it? Answer with a number.

name: _______________________________________________________ number:______________________________

10 (21)
Questions 14:

Consider the behavior tree below, designed for a mobile manipulator.

What actions would be performed in what order, if the robot was starting
in the following configurations (1p for each correct answer, note that
actions are numbered in the figure) Q14: a)....3,2,1...b)..no action.. c) 3,(2,1)
(2,1)(2,1) etc d).no actions

a) The object is not at the goal, and the robot is not near the
object.

b) The object is at the goal and the robot is not near the object.

c) The object is not at the goal, and the robot is not near the
object, but “Drop object at goal” has a bug that makes the
object end up slightly outside the goal area.

d) The object is at the goal, and the robot is not near the object,
but “Drop object at goal” has a bug that makes the object end
up slightly outside the goal area.

_
Question 15:

Given a behavior tree for playing a Mario type of computer game, given
sensor input in the form of a 5 by 5 grid of cells numbered from 0 to 24,
as indicated in the figure. What action would be immediately applied in
the following situations? (1p for each correct answer) Q15:
a)....shoot..b)....jump....... c)..........jump........d)...move right

name: _______________________________________________________ number:______________________________

11 (21)
 a) Enemy in Cell 14 and Obstacle in Cell 12 returns
Success
b) Enemy in Cell 13 and Obstacle in Cell 12 returns
Success
c) Not obstacle in Cell 16 and not obstacle in Cell 17
returns
Success
d) Not obstacle in Cell 16 returns Success

Question 16:

For each alternative below, mark if it as true or false, regarding a

differential drive robot. (2p per correctly classified statement.)

a) The name differential drive comes from the fact that you need to
have two different types of motors to implement a differential drive
robot. Since the motors are mounted in opposite directions, one
motor must be a forward moving motor and the other a reverse type
motor. F
b) If we put big wheels on the robot we typically want a gear ratio such
that every turn of the motor axis results in more rotations of the
wheel than when we use smaller wheels. F

Question 17:

name: _______________________________________________________ number:______________________________

12 (21)
Mark each of the following statements as true or false regarding legged
and wheeled robots. (4p if all are correct.)

a) Legs typically lead to a more complicated mechanical design.T

b) Legs typically lead to using less power when moving from A to B.F
c) Legs typically lead to a robot that requires more complicated
control.T

Question 18:

Mark each of the following statements as true or false regarding a

RGB-D sensors. (4p if both are correct.)

a) The ‘D’ stands for depth and the most common way to estimate this
information in consumer type sensors, like on more advanced
phones, is to measure the intensity of the light that hits each pixel.F
b) Before RGB-D sensors were available, a common way to acquire
depth information was to use two cameras and to calculate the
depth using triangulation.T

Question 19:

Consider the problem of localising a robot that moves in the plane and
has access to a compass that can tell the orientation accurately at all
times. We want to use a particle filter.

The particle filter is described by the following algorithm

0.1. Initialize the particlesand with weightsUse odometry to predict the

new state of each particle and perturbπ𝑖0 =𝑖 =1/{𝑁1., ., 𝑁} given what
you know to start with

each particle according to odometry noise.

name: _______________________________________________________ number:______________________________

13 (21)
 2. Multiply the weight,of the measurementπ𝑖𝑘 .of each particle 𝑖, with the
likelihood 𝑝(𝑧𝑘|𝑥𝑖𝑘)

3. Re-sample “if needed” and then return to 1.𝑧𝑘

We know that the robot starts near the recharging station and we decide
to initialize the state estimate with a uniform distribution in a circle with
radius 1m around the recharger position which is (0,0).

Which option is most correct for the two questions below?

(2p per correct answer.)

a) The state to be estimated by the filter is best represented as

( (5

b) We get an accurate measurement of the absolute position of

therobot that says the robot is in fact at x=3m, y=0m. Which of
the2)5)4)6)3)1) (((𝑥𝑥𝑧𝑥θ𝑥,,,,,)θ𝑦𝑦𝑦𝑦),,,)θ𝑧𝑧),)θ)

particle sets a-c below best represents the states of the particles at
the end of step 2 with this measurement?

name: _______________________________________________________ number:______________________________

14 (21)
 a.

b.

name: _______________________________________________________ number:______________________________

15 (21)
 c.

Question 20:
Consider a 1D system described by the equations

●● System dynamics:Measurement model:𝑥𝑘 =𝑧𝑘 𝑥=𝑘− 1 𝑥𝑘+ − ∆𝑡 𝐿· | |

𝑢+𝑘−1 𝑣 a landmark. The state is𝑘+ 𝑤𝑘−1

whereestimated in meters.We are using a Kalman filter to estimate the

position. The algorithm is𝐿 is the very accurately known position of

below for your reference.

Assume thatbelow as true𝑃or𝑘− false= 0. 1if we use a distance sensorfor

a give time step 𝑘. Mark each statementwith standard

name: _______________________________________________________ number:______________________________

16 (21)
deviation of about 1 mm to measure the distance to the known landmark.
(2p for each correct answer)

a) A measurement, at this time, will have a large impact on the state

estimate. T
b) The value of 𝑅𝑘 would be about 10−3. F

Question 21:

Consider a system described by the same equations as in Question 20

where we also use a Kalman Filter for state estimation. Assume the
following

● L = 0 (the landmark location, see Question 20)

● −
(the state estimate after the prediction step)

●● 𝑥𝑃𝑘− == 1−0001

𝑘 (the true state)

What option best describes how the change in the state estimate after

the𝑥𝑘 = 2 measurement update step with a measurement 𝑧𝑘 = 1. (4p)

1. It increases, i.e. moves to towards the true state

2. It is unchanged
3. It decreases, i.e. moves away from the true state

Question 22:

Mark each statement as true or false regarding occupancy grids.

name: _______________________________________________________ number:______________________________

17 (21)
 (2p per correctly classified statement.)

a) There are no spatial correlations in environments. Occupancy grid

cells can therefore be updated independently of each other. This is
great because it saves computations. F
b) In indoor office environments when using an occupancy grid for
local path planning it is almost always enough to have grid cells
that are of the same size as the robot. F

Question 23:

Let G be a 2D grid and let the value in each cell be calculated according
to the following pseudocode for processing a point cloud.

1. Associate each point in the point cloud with a cell based on its x,y
coordinate.
2. For all points that are associated with cell i:
a. Remove points that have a z-value larger than Z max.
b. Assign the value of cell i to be the max of the previous value and
the max of the z-values for the remaining points.

Which map type best matches this description? (4p for correct answer)

1) Feature map
2) Topological map 3) Elevation map
4) Occupancy map

Question 24:

Mark each statement regarding mapping and SLAM as either true or

false. (2p per correctly classified statement.)

a) In mapping the position is always known with extremely high

accuracy, typically in the range of fractions of a millimeter.F
b) In SLAM there is uncertainty in both the robot pose and the map.T

Question 25:

name: _______________________________________________________ number:______________________________

18 (21)
Make the best match between the type of dynamic information (a and b)
and where in the navigation system it is best dealt with (1 or 2).
(4p for correct answer)

Type of dynamics

a) You know the average flow of people in different places in a

building.1
b) You detect a person moving in front of the robot.2

Place in the navigation stack

1) Global motion planning

2) Local motion planning

name: _______________________________________________________ number:______________________________

19 (21)
name: _______________________________________________________ number:______________________________
21 (22)
Answer sheet

Q1: a).....6......... b)....4...... c)...3......d)....3..... (1 p for each correct answer).

Q2: a).....3......... b)....6...... (2p for each correct answer)
Q3: ………b c e…………. (4p for correct list, 2p if ONE missing OR added)

Q4: ……a b f g… (4p for correct list. 2p if one statement is missing. Else 0)
Q5: a)..F.......... b)...F...... c)..T........ d)...T....... (1p per correct answer.)
Q6: ………3………….. (The correct answer gives 4 p)
Q7: ……d a c b……………..(Complete list in correct order gives 4 p)
Q8: a)...P........... b).....P..... c)..P........ d)....P....... (1p per correct answer)
Q9: ……b d…………….. (4 pts. for correct list 2p if ONE missing OR added)
Q10: a).....T......... b)...F....... c)...T...... d)....F.... (1 p per correct answer)
Q11: a)....(7-12) and (14-16)..............(3 p for correct answer, 1p for one range)
b)..........Change “x_min” to “x_nearest" (1 p for correct answer)

Q12: a)...17±1 m.b).No.....c) 5 + 6 2 ±1 m d)..No... (1p per correct answer)

Q13: a).shorter.... b)..yes………c).....1 m… d).19 ±1 m.(1p per correct answer)

Q14: a)....3,2,1...b)..no action.. c) 3,(2,1)(2,1)(2,1) etc d).no actions (1p per c a)
Q15: a)....shoot..b)....jump....... c)..........jump........d)...move right (1p per c a)
Q16: a)......F.......... b).......F........... (2 p per correct answer)
Q17: a).......T......... b)......F............ c)...........T....... (4p if all are correct)
Q18: a)......F.......... b).........T......... (2p per correct answer)
Q19: a).......5........ b)........b.......... (2p per correct answer)
Q20: a).....T............. b).........F......... (2p per correct answer)
Q21: ......2.................... (4p for correct answer)
Q22: a)......F.......... b)........F.......... (2p per correct answer)
Q23: ...........3............... (4p for correct answer)
Q24: a).....F............. b).......T........... (2p per correct answer) Q25:
a)........1.......... b)......2............ (4p if both are correct)
name: _______________________________________________________ number:______________________________
22 (22)