Intelligence Mobile Robotics Homework 03

Intelligence Mobile Robotics
Homework 03

1 Task One
The main objective of this task is to get a proper understanding of EKF (Extended Kalman Filter)
localization. Design a relationship between the current and successive pose of differential drive
robot.
1. What is the state variables vector for the robot (xk)?
2. Design the system model (Φk)
3. A sensor that attached to the robot gives the distances to each of the visible landmarks. Hence,
range can be obtained to each sensor as follow. Assume, p
i
x
and p
i
y
are the distances on x and
y direction respectively. r =
q
(p
i
x − x)
2 + (p
i
y − y)
2 where i depicts i
th landmark. Relative
orientation to each landmark φ = arctan( p
i
y−y
p
i
x−x
) − θ. Using this information try to obtain
the measurement model (h(¯x
−
k
))? Here, it is assumed that current position of the robot as x
and y and heading of the robot as θ. You may change these two formulas adhere with your
design. Also, you may decide the positions of the landmarks based on your configuration
space. Only constrain is that you should have at least 3 landmarks.
4. Write down all the assumptions about errors of the filter?
5. The robot should route through two paths: straight line and taking a turn. Draw Pk changes
over time for each of the paths where at least 4 timestamps would be enough: start position,
end position and a few in between start and end position. Importance: EKF should be
implemented by yourself. You may use any programming language.
1
2 Task Two
Id Name
01 Sabirova Adelya
02 Ahmed Nawaz
03 Andrey Stepanov
04 Arslan Siddique
05 Aydar Ahmetzyanov
06 Dmitriy Desyatkin
07 Lyailya Aminova
08 Maksim Rassabin
09 Oleg Balakhnov
10 Sami Sellami
11 Valeriya Skvortsova
12 Victor Massague Respall
In hw-02, you’ve touched upon a step into a tracking problem especially into data association. Here
you’re going to move a step ahead and try to incorporate data association into tracking. Each of you
is provided with an image sequence of a scenario. After browsing your image set, try to seek an
object which is moving through a few images (at least 5 images) continuously. Locate a bounding
box for a chosen object which can be done either manually or automatically. Assuming an object
is not going away from the initial bounding box, try to find some feature points for the background
using any feature descriptor. Now assume those feature points are the landmarks for the robot you
have developed for the task one. Update EKF for as suited for this task. Assume robot is moving
on a straight line. If you want to make any additional assumptions please state them as well. Please
explain how you detected the feature points and why you decided so.
3 Submit
What should you turn in? Please, upload the single zip file which includes your source code (task
01 and task 02) and report for both tasks.