An improved Q-learning algorithm(ISTM)integrating spatio-temporal mechanism was pro-posed.According to the temporal and spatial variation characteristics of obstacles in the dynamic scene,a dynamic environment with mixed dynamic and static obstacles was established.The dynamic reward and punishment function was constructed by introducing the space-time mechanism,and the path search was carried out according to the dynamic change of obstacles,so as to improve the percep-tion ability of unmanned boats to the change of obstacle state.The dynamic exploration mechanism was established,and the greedy factor was adjusted adaptively to improve the exploration efficiency of unmanned boats in dynamic scenes.The results show that the path planning based on ISTM algorithm has shorter convergence time,higher stability and shorter planned path.
维普
万方数据