摘要
为避免深度神经网络路径决策失败给采摘机器人造成损害,提出一种适用于深度强化学习采摘机器人的安全防护设计方案。方案采用多组神经网络共同对机器人的状态进行决策,并利用V-Rep搭建平台进行仿真,结果表明,采用安全方案驱动机器人时,不仅能遏止机器人的错误动作,还能有效提高机器人路径决策的成功率。上述研究对设计深度强化学习机器人控制方案有一定的参考价值。
In order to avoid damage to the picking robot due to the failure of the deep neural network path decision,this paper proposes a safety protection design scheme for the deep reinforcement learning picking robot.The program uses multiple sets of neural networks to jointly make decisions on the state of the robot,and uses V-Rep to build a platform for simulation experiments.The experimental results show that when the safety system is used to drive the robot,it can not only prevent the robot's wrong actions,but also effectively improve the success rate of the robot's path decision.This research has certain reference value for the design of deep reinforcement learning robot control scheme.
作者
赵锦泽
王好臣
于跃华
李家鹏
ZHAO Jin-ze;WANG Hao-chen;YU Yue-hua;LI Jia-peng(School of Mechanical Engineering,Shandong University of Technology,Zibo Shandong 255000,China)
出处
《计算机仿真》
北大核心
2023年第5期448-453,共6页
Computer Simulation
基金
淄博市重点研发计划项目(2019ZBXC585)。
关键词
采摘机器人
深度强化学习
安全算法
路径规划
机器人正解
Picking robot
Deep reinforcement learning
Security algorithm
Path planning
Robot forward solution