基于自适应熵值的移动机器人无地图导航方法OA

Aiming at the problem of poor ability to escape from local optima after a robot became trapped,which existed in existing mapless navigation methods based on machine learning,an adaptive entropy mapless navigation method was proposed.First,this method determined whether a local optimum phenomenon occurred according to the reward fluctuations during the robot's training process,and then on this basis,continued to improve the entropy adjustment mechanism in the soft actor-critic(SAC)algorithm to achieve the effect of adaptively changing the temperature coefficient to strengthen its learning ability after the robot became trapped in a local optimum,and finally promoted the robot to successfully learn a globally optimal navigation strategy.Simulation results show that compared with existing methods,the navigation strategy trained by the proposed method has stronger escape ability and higher navigation success rate.