In this thesis, an existing odometry system of a line-following mobile robot was upgraded
with a real-time position-calibration system. The robot tracks its position along the track as
the travelled distance estimated from the wheel encoders, so the position error accumulates
lap after lap due to wheel slip, encoder quantization and integration error of the heading
angle. The developed system corrects this error periodically: a Python program segments the
reference lap, determines reset points and builds reference windows of heading-angle
changes, while a program running on the robot's microcontroller compares the measured
pattern of heading-angle changes with the reference using the sum of squared differences
and resets the travelled distance. Sampling is bound to the travelled distance, which makes
the matching independent of the driving speed. The system was evaluated experimentally on
two tracks at five driving speeds by manually measuring the deviation after each of 20
consecutive laps. Without calibration the deviation reached up to 3150 mm after 20 laps,
whereas with calibration it remained bounded within 157 mm regardless of the number of
laps.
|
