一种无先验地图的移动机器人导航方法

为了提高移动机器人在复杂环境下的无先验地图导航算法模型训练速度及导航成功率，提出一种基于深度确定性策略梯度(deep deterministic policy gradient，DDPG)的移动机器人导航方法。利用2D激光雷达的均匀分布测距信息，降低环境噪声的干扰及高维度环境信息的计算量;采用人工势场法构建移动机器人从初始位置到目标点过程的奖励函数;通过Actor-Critic网络结构提高模型训练的稳定性和泛化能力。实验结果表明，提出的方法具有模型训练速度快、导航成功率高及泛化能力强等优点。

A mapless navigation method for the mobile robot

In order to improve the training speed and navigation success rate of mobile robot in unknown complex environment without prior map navigation algorithm model, a navigation decision method of mobile robot based on DDPG are proposed in this paper. In this method, the uniform distribution of 2D laser radar ranging information is extracted to reduce environmental noise interference and high dimension computation of environment information, the artificial potential field method is used to construct the reward function of the process from the initial position to the target point, encourage the robot to move to the target position and avoid obstacles. A network structure based on Actor-Critic is adopted to strengthen the stability and generalization ability of model training. Experimental results show that the proposed navigation model has the advantages of fast training speed, high navigation success rate, and strong generalization ability.