Robotics Landmark Navigation Using Image and Sonar Processing

1
Robotics Landmark Navigation Using Image and
Sonar Processing

• Andrea Fey
• April 20, 2006

2
Robot Simulation

• Simulated Pioneer robot using landmarks to map
  its environment.

sonar sensors
camera field of view indicator
landmarks
3
Sensors

• Camera view
• Sonar query returns distances
• gtgtgtself.robot.sonar0.distances()
• 6.7452803435804594, 6.4821348706844084,
  4.7049387876403461, 3.7200875453266398,
  3.9169508041779628, 3.8103321040585008,
  3.1074810780405624, 3.0402629280589757

4
Dead Reckoning

• Robot calculates its position based on number of
  times each of its two wheels rotates
• Technique degrades over time

5
Landmark Navigation

• Robot identifies a landmark and determines
  position relative to it
• Robot knows the location of the landmark

6
Calculation of Position
x
ß'
?
a
ß
?b
?a
y
h
c
b
Formulae a2 b2 2abcos( ?a - ?b ) c2 a2
c2 2accos( ß ) b2 h asin( ß ) ? -?a ß
- 90º
7
Comparison of Methods

• Results of two robots running for forty-five
  minutes in a square pattern. The landmark
  navigator knows reasonably well where it is in
  the environment, but the dead-reckoning robot is
  inaccurate.

8
Mapping the Environment

• I chose to map the ten-by-ten meter environment
  to a ten-by-ten grid (plus two rows/columns for
  the outer walls)
• Decreased grid count if there was a wall
  increased if free space
• Initialized known locations to large negative
  numbers to reduce error

9
Bresenham
Image from Wikipedia

• Bresenhams line drawing algorithm determines the
  grid locations between the robot and the sonar
  hit
• Computationally cheap integer arithmetic

10
Environment Map
11
Learnings

• Linux administration
• Installation
• Creating accounts
• Grub, RPM, yum
• Python
• Modules
• Lists
• Robotics
• Sensor processing
• Robot mechanics
• New application of Bresenhams line drawing
  algorithm
• Using software with limited documentation