| 摘 要: 为了解决现有的机械臂焊接系统调整动作的难度大，缺乏灵活性的问题，本文采用了深度强化学习算法 来解决机械臂的路径规划问题；该方法使用一个三层的DNN网络，输入为机械臂的状态信息，输出为机械臂的运动关 节角度，通过离线训练，机械臂能够自行训练出一条接近于最优的运动轨迹，能够成功地避开障碍物到达目标点；仿真 在一个三自由度点焊机器人的模拟平台上进行，仿真实验表明，采用深度强化学习技术的机械臂能为焊接机械臂规划出 一条无碰撞的路径，具有较强的避障能力。 |
| 关键词: 机械臂；强化学习；碰撞检测；路径规划 |
|
中图分类号: TP391.9
文献标识码: A
|
| 基金项目: 本文系广东省应用型科技研发专项基金项目“复杂构件的智能焊接机器人研发与产业化”(项目编号：2015B090922013). |
|
| Research on Obstacle Avoidance Path Planning of the Mechanical Arm Based on Deep Reinforcement Learning |
|
LI Guangchuang,CHEN Lianglun
|
( College of Automation, Guangdong University of Technology, Guangzhou 525000, China)
|
| Abstract: In order to solve the difficulty of adjusting the existing mechanical arm welding system and the lack of flexibility,this paper adopts a deep reinforcement learning algorithm to solve the path planning problem of the mechanical arm;the method uses a three-layer DNN network with the input of mechanical state information and the output of the joint angle of the arm.Through offline training,the mechanical arm can be trained to move a nearly optimal motion trajectory and successfully avoid obstacles to reach the target point;the simulation is carried out in a three-degree-of-freedom welding robot simulation platform.The simulation experiment shows that the mechanical arm with deep reinforcement learning technology can plan a collision-free path for the welding mechanical arm and has strong obstacle avoidance ability |
| Keywords: mechanical arm;reinforcement learning;collision detection;path planning |
