融合DDPG算法的移动机器人路径规划研究

Path Planning of Mobile Robot with Fusion DDPG Algorithm

路径规划是实现移动机器人自主导航的关键技术。针对传统算法不能有效解决未知动态环境下移动机器人路径规划的问题,提出了一种基于深度确定性策略梯度算法(DDPG)与人工势场法相融合的路径规划方法,首先创建并训练基于DDPG算法的路径规划模型;其次,利用人工势场法对DDPG算法的动作选择策略进行干预;最后,在四种仿真环境下验证本文所提算法的性能。仿真实验结果表明,本文提出的算法与DDPG算法相比,移动机器人行驶路径长度减少3%至10%,行驶过程中角速度超过1.0rad/s的次数减少5%至12.5%,表明所提方法能有效提升规划路径的平滑度,同时缩短移动机器人路径规划长度。此外,相较人工势场法,本文所提融合算法可有效解决未知动态环境下移动机器人路径规划难题。

Path planning is one of the vital technologies to realize autonomous navigation of mobile robot.Aiming at the problem that the traditional algorithm cannot effectively achieve path planning of mobile robot in an unknown dynamic environment, a fusion path planning algorithm based on deep deterministic policy gradient(DDPG) algorithm and artificial potential field method is proposed in this paper. Firstly, the path planning model based on DDPG algorithm is created and trained. Secondly, artificial potential field method is used to intervene the action selection strategy of DDPG algorithn. Finally, the performance of the proposed fusion algorithm is verified in four different simulation environments. The simulation results show that the proposed fusion algorithm reduces the path length of mobile robot by 3% to 10%, and the times of the angular velocity exceeding 1.0 rad/s during driving is reduced by 5% to 12.5% compared with DDPG algorithm. The proposed fusion DDPG algorithm can effectively improve the smoothness of the planned path, and also shorten the path. In addition, the proposed algorithm can also achieve path planning of mobile robot in an unknown dynamic environment compared with artificial potential field method.