Neural Network Reactive Navigation and Control of Wheeled Mobile Robot

Abstract

A neural net real-time obstacle avoidance and control approach for mobile robot has been developed and numerically implemented. A collision-free path is calculated using an efficient neural net motion planer. The output of the navigation level is fed into a neural net tracking controller that takes into account the complete dynamics of the mobile robot. The proposed neural reactive navigation approach is based on the coordination of elementary behaviors. To avoid the convex obstacles the neural navigator fuses a ‘reaching the middle of a collision-free space’ behavior and a ‘goal-seeking’ behavior. A ‘wall-following’ behavior was conducted too, which can be applied to avoid the concave obstacles. The structure of the neural-net controller for nonholonomic mobile robot is derived using a filtered error approach. The effectiveness of the proposed method is numerically verified by a series of experiments on the emulator of wheeled mobile robot Pioneer-2DX.