Accurate parking of nonholonomic robots

Abstract

Due to nonholonomic constraints as well as constraints on torque and power resources, it is challenging to develop an auto-parking system for nonholonomic robots to asymptotically stabilize the robots at their goal pose. Many existing controllers escape from the singularity set which is believed to be uncontrollable. Recently, a new type of parking algorithm was proposed in [11]. Instead of escaping from the singularity set, the proposed controller will approach it [11]. The efficiency of the controller is only analysed and verified via Matlab simulations. This objective of this final year project is to focus on the implementation of the new parking algorithm on a robot using the Robot Operating System (ROS). The fusion of two localisation techniques using Adaptive Monte Carlo Localisation (AMCL) and Map-Based Odometry is effective in the testing of the new algorithm. The new parking algorithm can achieve a shorter path and reduced control effort as compared to the existing method. With its simplicity and efficiency, our implementation has a potential to be used in practice, such as parking of wheelchairs and charging of vacuum cleaners. The report begins with an introduction to the parking of nonholonomic robot. Existing methods will be discussed in the literature review. The remaining chapters will present the simulation and implementation results on the robot. The report will conclude with a summary and prospective plan for future work.