Neural-fuzzy hybrid system for mobile robot path-planning in a partially known environment

In this paper, a neural-fuzzy hybrid control approach is proposed for controlling a mobile robot that can avoid an unexpected obstacle in a navigational space. First, to describe the global structure of a known environment, a heuristic dominant collision-free space band is introduced. Based on the band, the moving information in the known environment is trained to a neural controller. Then, during the execution of a mobile robot navigation moving information at each position is given from the neural controller. If the mobile robot encounters an unexpected obstacle, a fuzzy controller is activated to adjust the moving information given from the neural controller to avoid the unexpected obstacle. When the robot has safely avoided the obstacle and resume the originally taught path, the fuzzy controller is deactivated. Some numerical examples are presented to demonstrate the planning algorithm