基于粒子群RBF神经网络的双关节机械臂系统控制  

Control of Dual-Joint Manipulator System Based on Particle Swarm RBF Neural Network

作　　者：郑明军 兰庆洋 吴文江 Zheng Mingjun;Lan Qingyang;Wu Wenjiang(School of Mechanical Engineering,Shijiazhuang Tiedao University,Shijiazhuang 050043,China;Office of Academic Affairs,Shijiazhuang Tiedao University,Shijiazhuang 050043,China)

机构地区：[1]石家庄铁道大学机械工程学院,河北石家庄050043 [2]石家庄铁道大学教务处,河北石家庄050043

出　　处：《石家庄铁道大学学报（自然科学版）》2021年第4期46-52,共7页Journal of Shijiazhuang Tiedao University(Natural Science Edition)

基　　金：河北省自然科学基金(E2017210166);河北省教育厅资助科研项目(ZD2020320)。

摘　　要：针对RBF神经网络算法用于控制时难以求解网络隐含层参数中心向量c和标准化常数b的问题,提出基于粒子群参数优化的RBF神经网络(PSO-RBF神经网络)控制方法。建立旅客列车自动上水装置双关节机械臂动力学模型,将粒子群算法与RBF神经网络控制机械臂动力学特性结合,在连续空间快速搜索网络隐含层参数最优解,得到PSO-RBF神经网络控制方法;建立针对双关节机械臂的PSO-RBF神经网络控制系统并进行仿真,与基于遗传算法调节隐含层参数的RBF神经网络控制方法进行对比和分析。研究表明,采用PSO-RBF神经网络控制方法可以有效避免机械臂控制失效,能够使肩关节和肘关节响应时间缩短52%和47%,最大稳态误差减小49%和58%,平均稳态误差减小54%和55%。Aiming at the problem that it is difficult to obtain the center vector c and the standardization constant b of the hidden layer parameters when RBF neural network algorithm is used for control,a RBF neural network control method based on particle swarm optimization(PSO-RBF neural network)was proposed.A dynamic model of a double-joint mechanical arm of an automatic water feeding device of a passenger train was established,and a PSO-RBF neural network control method was obtained by combining a particle swarm optimization algorithm and a RBF neural network to control the dynamic characteristics of the mechanical arm and rapidly searching the optimal solution of network hidden layer parameters in a continuous space.The PSO-RBF neural network control system for the dual-joint manipulator was established and simulated,and compared with the RBF neural network control method based on genetic algorithm to adjust the hidden layer parameters.The results show that the PSO-RBF neural network control method can effectively avoid the control failure of the manipulator,and can reduce the response time of the shoulder and elbow joints by 52%and 47%,the maximum steady-state error by 49%and 58%,and the average steady-state error by 54%and 55%.

关 键 词：旅客列车自动上水装置 双关节机械臂 径向基神经网络 粒子群算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]