Adaptive Control Simulation of AUV Using Neural Controllers Trained by GA

Abstract

In this research, we propose a method to adaptively generate AUV trajectories using a neural controller that learns by applying a genetic algorithm. The method uses two neuro-controllers. One is learned in advance as a controller. The other is learning at the same time as redesigning a new trajectory when the AUV deviates from the planned trajectory during navigation. The two neural controllers are evolutionarily learned using genetic algorithms. In the navigation simulation, the multi-point search problem of AUV under the tidal current environment was taken up. Numerical experiments have shown the effectiveness of the proposed method.