基于仿人自适应DDPG算法的机械臂运动规划研究  

Research on motion planning of manipulator based on humanoid adaptive DDPG algorithm

作　　者：汪红[1] 解伟 田莎莎[1] 帖军[1] 昂寅 WANG Hong;XIE Wei;TIAN Shasha;TIE Jun;ANG Yin(College of Computer Science&Hubei Provincial Engineering Research Center of Agricultural Blockchain and Intelligent Management&Hubei Provincial Engineering Research Center for Intelligent Management of Manufacturing Enterprises,South-Central Minzu University,Wuhan,430074,China)

机构地区：[1]中南民族大学计算机科学学院&农业区块链与智能管理湖北省工程研究中心&湖北省制造企业智能管理工程技术研究中心,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2023年第3期334-342,共9页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家民委中青年英才培养计划(MZR20007);湖北省科技重大专项(2020AEA011);武汉市科技计划应用基础前沿项目(2020020601012267);中央高校基本科研业务费专项资金资助项目(CZQ21026)。

摘　　要：现代大棚高架栽培技术中作物种植密度大,机械臂采摘容易受到旁边植株影响.针对这一问题,将人手臂的运动约束应用到机械臂的DDPG运动规划算法中,从而让机械臂具有人手臂的灵活性.所提出的仿人自适应DDPG算法新增自适应采样策略来对经验缓冲池中轨迹经验的优先级进行自适应动态调整,从而避免算法陷入局部最优,加快算法的训练速度.将仿人自适应DDPG算法在PyBullet仿真平台上进行实验,经过300轮训练,最终机械臂的抓取成功率高达96.5%,且机械臂在执行抓取操作的过程中各关节的运动范围皆符合人手臂运动时的关节范围,证明机械臂具有人手臂的运动特性,能更好地应用于大棚高架栽培作物的采摘.In the modern greenhouse elevated cultivation technology,the crop planting density is high,and the picking of the robotic arm is easily affected by the plants next to it.In response to this problem,the motion constraints of the human arm is applied to the DDPG motion planning algorithm of the robotic arm,so that the robotic arm has the flexibility of a human arm.The human-like adaptive DDPG algorithm proposed adds an adaptive sampling strategy to dynamically adjust the priority of the trajectory experience in the experience buffer pool,so as to avoid the algorithm from falling into local optimum and speed up the training speed of the algorithm.The humanoid adaptive DDPG algorithm was tested on the PyBullet simulation platform.After 300 rounds of training,the final grasping success rate of the robotic arm was as high as 96.5%,and the range of motion of each joint during the grasping operation of the robotic arm was consistent with the joint range of the human arm movement proves that the robotic arm has the movement characteristics of the human arm,and can be better applied to the picking of elevated cultivation crops in greenhouses.

关 键 词：机械臂 奖励函数 运动约束 自适应 

分 类 号：O625.67[理学—有机化学] O643.3[理学—化学]