Actor-Critic框架下基于DDPG算法的绘画机器人控制系统优化设计  

Optimization Design of Painting Robot Control System Based on DDPG Algorithm under Actor-Critic Framework

作　　者：罗子彪 唐娇 LUO Zibiao;TANG Jiao(Guangxi Vocational College of Industry,Nanning 530000,China;Guigang city people’s hospital,Guigang,Guangxi 537100,China)

机构地区：[1]广西工业职业技术学院,南宁530000 [2]贵港市人民医院,广西贵港537100

出　　处：《自动化与仪器仪表》2025年第2期193-197,202,共6页Automation & Instrumentation

摘　　要：人工智能与艺术创作的碰撞成为当前研究新焦点。然而,机器人在进行图画绘制工作中的控制效果却难以满足精度需求。因此,研究在深度确定性策略梯度算法基础上进行了绘画机器人控制系统设计。在Actor网络和Critic网络框架下,对算法的奖励函数以及经验池进行改进与优化,并提出了绘画机器人控制系统。验证显示,研究提出的控制系统比其他算法基础上的控制系统训练收敛速度平均提高了38.04%。机械臂肘关节仿真误差比其他算法平均减少了93.74%。结果表明,对算法的奖励函数与经验池进行改进能够提高算法收敛速度与性能。研究提出的绘画机器人控制系统对机器人绘制图像的过程控制能够满足控制精度需求,在机器人控制中具有积极的应用价值。The collision between artificial intelligence and art creation has become a new focus of current research.However,the control effect of the robot in performing drawing work is difficult to meet the accuracy requirements.Therefore,the research carried out the drawing robot control system design based on the deep deterministic policy gradient algorithm.Under the framework of Actor network and Critic network,the reward function as well as the experience pool of the algorithm are improved and optimized,and the drawing robot control system is proposed.The validation shows that the proposed control system of the study shows an average increase of 38.04%in the convergence speed of the training of the control system over other algorithms based on the control system.The simu-lation error of the robotic arm elbow joint is reduced by 93.74%on average than other algorithms.The results show that improvements to the algorithm’s reward functions and experience pools can improve the algorithm's convergence speed and performance.The draw-ing robot control system proposed in the study can meet the control accuracy requirements for the process control of robot drawing ima-ges and has positive application value in robot control.

关 键 词：Actor网络 Critic网络 DDPG算法 深度强化学习 控制系统 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]