Path Tracking Control by Online Learning using CMAC for an Omnidirectional Mobile Robots

Abstract

In this study, we configured a path tracing control system that incorporated online learning using CMACs. The system has excellent performance for applying learning to general purposes, such as for omnidirectional mobile robots. Though CMAC has already been applied to process control and control of articulated robots, we adapted it for controlling mobile robots in this study. Generally speaking, traveling bodies must have high mobility in a narrow environment. An omnidirectional mobile robot with excellent traveling performance was used as the actual machine. The robot performed online learning using CMAC, a kind of neural network. In this paper, we propose an adaptive learning control system to implement path tracking control of mobile robots that travel with online input of the target value and that use the result of learning as the output value. A visual servoing system is used for this path tracking control system. We also report the results of various experiments on traveling by online learning. During the experiments on traveling, we were able to obtain excellent results from online learning and anticipate that teleoperation technology combining CMAC online leaning and a visual servoing system will be applied to care-taking devices and other systems.