Sidney N. Givigi Jr., James C. Lindsay, and Dany Ouellet
Autonomous robots, Fuzzy controller, Reinforcement learning
In this paper, we introduce a hybrid architecture based on a static PD controller and an adaptive fuzzy controller for the path tracking of an autonomous robot. The objective is to have the robot following a path generated by a path planning module. The PD controller is used as a first approximation of a tracker for the robot. However, as the dynamics of the robot changes and/or the path configuration changes and also when noise is present in the environment, an adaptive online fuzzy controller based on reinforcement learning is used to correct and improve the performance of the PD controller. Simulations demonstrate that the derived controller succeeds in providing a better performance for the system.
Important Links:
Go Back
