Calibration for Kinematic Control of Differential-Drive Mobile Robots: A Machine Learning Approach

• Main Authors: Zhuang, Yu-Hsiang, Yuan, Cheng-Zong, Lin, Chang-Yi, Peconkova, Veronika, Lee, Min-Fan Ricky
• Format: Conference Proceeding
• Language: English
• Subjects: Calibration; Force; kinematic control; Kinematics; Machine learning; mobile robot; Odometers; Odometry; Robustness
• ISSN: 2572-6919
• DOI: 10.1109/ARIS59192.2023.10268540

Odometer calibration is a crucial aspect of mobile robotics as it significantly impacts the robustness and accuracy of robot movement and positioning, while being a cost-effective approach. The primary goal of odometry calibration is to precisely adjust the robot's kinematic parameters. This is based on the premise that different movement patterns can lead to varying force states during motion. In this study, the robot was directed to traverse four distinct paths, with the third path being a combination of the first and second paths, and the fourth path being a refined version of the third. The performance of combined parameters for the third and fourth paths was compared to using single parameters for each path. The results demonstrated a significant improvement in performance for both the combined and single parameters when compared to the original parameters, with the most substantial enhancement observed in the error of the single parameters. These findings suggest that the utilization of combined parameters may not necessarily be inferior to the use of single parameters, though further comprehensive testing is required to fully evaluate the benefits of this approach.