EE 468/CE 468: Mobile Robotics Fall 2022

Habib University
Course Code: EE 468/CE 468: Mobile Robotics
Course Title: Mobile Robotics
Instructor name: Dr. Basit Memon
Examination: Quiz 2 Exam Date: November 19, 2022
Total Marks: 100 Duration: 50 minutes

Instructions
1. You can refer to course slides, your notes, homework solutions, or any of the books
included in the course syllabus. You’re not permitted to specifically search for responses
to any of the exam questions online. Where appropriate, you can cite the slides and
don’t have to redo what has already been done.
2. The exam will be administered under HU student code of conduct (see Chapter 3 of
https://habibuniversity.sharepoint.com/sites/Student/code-of-conduct). You
may not talk, discuss, compare, copy from, or consult with any other human on this
planet (except me) on questions or your responses during the exam.
3. Make sure that you show your work (intermediate steps). Most of the points for any
question will be given based on the followed process.
4. Kindly use a black/blue pen to write your hand-written solutions, so that they are legible.

Questions
Consider a robot that resides in a circular world consisting of ten different places that are Problem 1
numbered counterclockwise. The robot is unable to sense the number of its present place 85 points
directly. However, places 0, 3, and 6 contain a distinct landmark, whereas all other places do
not. All three of these landmarks look alike. The likelihood that the robot observes the land-
mark given it is in one of these places is 0.8. For all other places, the likelihood of observing
the landmark is 0.4. For each place on the circle compute the probability that the robot is in
that place given that the following sequence of actions is carried out deterministically and the
following sequence of observations is obtained: The robot detects a landmark, moves 3 grid
cells counterclockwise and detects a landmark, and then moves 4 grid cells counterclockwise
and finally perceives no landmark.

We can use Markov grid-based localization here. Let xk be the location of the robot Solution 1
at step k.
xk ∈ {0, 1, · · · , 9}.

©Basit Memon 1
EE 468/CE 468: Mobile Robotics Fall 2022

Initially, we have no information about the place of the robot. So the prior belief is:

1
bel(x0 = i ) = , ∀i ∈ {0, 1, · · · , 9}.
10

The measurement at step k is denoted by zk ∈ {0, 1}, where zk = 1 indicates that a landmark
was detected. We’re provided that
{
0.8 if xk ∈ {0, 3, 6}
p(zk = 1|xk ) =
0.4 otherwise.

Correspondingly,
{
0.2 if xk ∈ {0, 3, 6}
p(zk = 0|xk ) =
0.6 otherwise.

Our motion model is deterministic. If u is the number of steps, then our motion model can
be described as:
{
1 if xk = xk−1 + uk (mod 10)
p(xk |xk−1 , uk ) =
0 otherwise.

Let’s iterate through the measurement and actions sequences step by step. Given that z0 = 1,

bel(x0 ) = ηp(z0 |x0 )bel(x0 )

{
0.8/10 if x0 ∈ {0, 3, 6}
=η
0.4/10 otherwise.
∑
⇒ η = 1/ bel(x0 ) = 1/0.52
x0

Next, the robot has moved three steps counter-clockwise. So, our predicted belief at next
time step is
∑
bel(x1 ) = p(x1 |x0 , u1 = 3)bel(x0 )
x0
= bel(x0 = x1 − 3)
{
0.154 if x1 ∈ {3, 6, 9}
=
0.077 otherwise.

©Basit Memon 2
EE 468/CE 468: Mobile Robotics Fall 2022

A landmark is detected. So, we have

bel(x1 ) = ηp(z1 = 1|x1 ) bel(x1 )

({ ) ({ )
0.8 if x1 ∈ {0, 3, 6} 0.154 if x1 ∈ {3, 6, 9}
=η
0.4 otherwise. 0.077 otherwise.

 0.123 if x1 ∈ {3, 6}
=η 0.062 if x1 ∈ {0, 9}

0.031 otherwise.
∑
⇒ η = 1/ bel(x0 ) = 1/0.56
x0

We move four grid cells counterclockwise. So, our predicted belief at next time step is
∑
bel(x2 ) = p(x2 |x1 , u1 = 4)bel(x1 )
x1
= bel(x1 = x2 − 4)

 0.22 if x2 ∈ {7, 0}
= 0.111 if x2 ∈ {4, 3}

0.055 otherwise.

A landmark is not detected. So, we have

bel(x2 ) = ηp(z2 = 0|x2 ) bel(x2 )

 
({ )  0.22 if x2 ∈ {7, 0}
0.2 if x2 ∈ {0, 3, 6} 
=η 0.111 if x2 ∈ {4, 3} 
0.6 otherwise. 
0.055 otherwise.


 0.044 if x2 = 0



 0.022 if x2 = 3

0.067 if x2 = 4
=η

 0.011 if x2 = 6



 0.132 if x2 = 7

0.033 otherwise.
∑
⇒ η = 1/ bel(x0 ) = 1/0.441
x0

Consider the polar to Cartesian coordinate transformations: x = r cos(θ) and y = r sin(θ). Problem 2
Assume that r and θ are corrupted by additive Gaussian noise, ϵr and ϵθ with a zero mean 15 points
and standard deviations of 0.01 meters and 0.3 rads respectively. It is evident that the vector
[ ]T
ρ = r θ is normally distributed.

©Basit Memon 3
 EE 468/CE 468: Mobile Robotics Fall 2022

[ ]T
(a) Is the vector α = x y also normally distributed? Justify.

(b) Approximate the mean of the vector α, when reference point is r = 1 and θ = 0.

(c) Assume that r and θ are uncorrelated. Given the covariance of the vector ρ, approximate
the covariance matrix of the vector α.

Solution 2 Using Slide 30-35 from Deck 08,

(a) No, in general α will not be normally distributed as transformation is nonlinear.

[ ]
r cos(θ)
(b) Let α = f (r, θ), where f = .
r sin(θ)

µα = f (1, 0)
[ ]
1
=
0

(c) The covariance matrix for ρ is

[ ]
(0.01)2 0
Σρ = .
0 (0.3)2

The covariance for α is approximately,

Σα = JΣρ J T

where
[ ∂f ∂f
]
J= ∂r ∂θ
[ ]
cos θ −r sin θ
=
sin θ r cos θ (r,θ)=(1,0)
[ ]
1 0
=
0 1
[ ][ ][ ]T
1 0 (0.01)2 0 1 0
⇒ Σα =
0 1 0 (0.3)2 0 1
[ ]
1 0
=
0 1

©Basit Memon 4