Robot Localization with Fiducial Markers
Name
Kristjan Laht
Abstract
Non-industrial robotics is a relatively new field with great potential for growth, but there are still many active research problems that prevent it from becoming ubiquitous. A robot not getting lost is one of those problems. Robots tracking their position based on wheel movement are subject to drift due to uneven surfaces and wheel slippage. One way for robots to determine their position and orientation (pose) without knowledge of its? drift is to use QR code like printable tags called fiducial markers that are simple and fast to recognize from images. These tags must be installed in the environment beforehand. This work integrates global pose from fiducial markers and local pose from wheel rotations into one package to create a system that accounts for the weaknesses of both. A simple environment where the use of this package was successful and a complicated environment where it failed are demonstrated. Fiducial markers used for finding the current pose of the robot are robust solutions in applicable environments, but improvements to robustness are still needed.
Graduation Thesis language
English
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Karl Kruusamäe
Defence year
2022