摘要
针对混联机械臂轨迹控制偏差大、难度高的问题,提出一种基于深度学习的容错算法。在笛卡尔空间内基于非线性函数确定关节向量间的相关性,并改善轨迹容错纠偏的收敛性能;基于DBNs模型训练混联机械臂动态位置移动信息,在全局范围内搜索最优解;根据每个关节空间移动轨迹特征和深度学习容错机制,保证机械臂系统闭环操作的稳定性和适用性。仿真结果表明:采用提出的容错算法,6个关节实际移动曲线与期望曲线的偏差程度较小,算法的复杂程度更低。
Aiming at the problems of large deviation and high difficulty in trajectory control of hybrid manipulator, a fault-tolerant algorithm based on depth learning was proposed. In Cartesian space, the correlation between joint vectors was determined based on non-linear function, and the convergence performance of the trajectory fault-tolerant correction was improved;the dynamic position movement information of the hybrid manipulator was trained based on DBNs model, and the optimal solution was searched in the global scope;the stability and applicability of the closed-loop operation of the manipulator system were guaranteed according to the trajectory characteristics of each joint space and the depth learning fault-tolerant mechanism. The simulation results show that the deviation between the actual movement curve and the expected curve of the six joints is small, and the complexity of the algorithm is lower.
作者
赵丽丽
ZHAO Lili(College of Information Engineering,Shandong Management University,Jinan Shandong 250357,China;Key Laboratory of TCM Data Cloud Service in Universities of Shandong(Shandong Management University),Jinan Shandong 250357,China)
出处
《机床与液压》
北大核心
2021年第3期35-40,共6页
Machine Tool & Hydraulics
基金
山东管理学院校级重点课题(XJ20180104)
国家中医药管理局中医药行业科研专项(2015468003-2-3)。
关键词
混联机械臂
深度学习
DBNs模型
收敛性能
容错机制
Hybrid manipulator
Deep learning
DBNs model
Convergence performance
Fault tolerance mechanism
