机电工程技术2024，Vol.53Issue(11)：96-100,5.DOI:10.3969/j.issn.1009-9492.2024.11.021
基于PPO算法的机械臂抓取策略研究
Research on Robotic Arm Grasping Strategy Based on PPO Algorithm
孔凡国 1仇展明 1王鑫 1陈靖轩 1袁功兴1
作者信息
- 1. 五邑大学机械与自动化工程学院,广东 江门 529020
- 折叠
摘要
Abstract
Aiming to address the path planning challenges faced by robotic arms when executing grasping tasks in multi-target scenarios,the feasibility of the Proximal Policy Optimization(PPO)algorithm in robotic arm grasping strategies is validated.The deep reinforcement learning based on the PPO algorithm is utilized,conducting interactive simulations of the robotic arm and objects in the PyBullet simulation environment.The positional coordinates of the robotic arm's end-effector and the target object serve as inputs to the deep reinforcement learning network,which outputs the next step's coordinates for the robotic arm's end-effector.Subsequently,the inverse kinematics of the robotic arm are employed to determine the rotation angles of its various joints as the action output.Finally,by integrating an optimized reward function,the study aims to enhance the learning efficiency of the robotic arm in training for grasping tasks and accelerate the convergence speed.Simulation experiments indicate that the application of the optimized reward function and the PPO algorithm enables the obtained reward values during robotic arm training to converge to a range of-50 to 0 around the 1000th iteration.Moreover,the robotic arm consistently completes grasping tasks within approximately 10 steps,validating the feasibility of the proposed method.The research achievement in robotic arm grasping tasks represents significant progress and provides robust methods and technological support for addressing complex grasping problems in practical applications.关键词
深度强化学习/PyBullet/机械臂抓取/PPOKey words
deep reinforcement learning/PyBullet/robotic arm grasping/PPO分类
信息技术与安全科学引用本文复制引用
孔凡国,仇展明,王鑫,陈靖轩,袁功兴..基于PPO算法的机械臂抓取策略研究[J].机电工程技术,2024,53(11):96-100,5.
