基于改进强化学习的移动机器人最短路径寻找方法OA

Path planning is a pivotal concern in mobile robotics,with effective planning significantly enhancing the operational efficiency of robots.The crux of the issue lies in accelerating the environ-mental exploration speed of mobile robots and identifying the shortest possible path.Current rein-forcement learning-based methods for shortest path discovery suffer from slow update speeds,resul-ting in excessive time spent on initialization and data iteration,particularly in smaller models.To address this,a novel method for shortest path discovery in mobile robots were proposed,leveraging reinforce-ment learning.Given the unique characteristics of unknown environments,the grid method was adap-ted and a strategy that simultaneously explores,interacts,and models was implemented.Further-more,a back tracking update strategy into traditional Q-learning to expedite convergence speed was incorporated.The experimental results demonstrate that when the ε-greedy algorithm and normaliza-tion methods are employed as update strategies,the proposed method significantly reduces the train-ing time required for robot pathfinding,while also enhancing training accuracy.The efficacy and ad-vantages of the proposed scheme are further validated by simulation results.