液压并联机器人的动态神经网络控制研究

Dynamic Neural Network Control of Hydraulic Parallel Robot

针对液压并联机器人运动过程中各单缸伺服系统的参数时变和耦合力扰动问题,利用在前馈型网络增加反馈环节的方法,设计了一种新型动态神经网络,并将该动态网络作为智能控制器应用于单缸伺服系统,同时根据能全面衡量系统性能的综合目标函数,推导出网络控制学习算法.仿真及试验结果表明,这种控制器的设计不依赖于系统模型,对于单缸系统的参数时变具有自适应性,对于耦合力扰动具有强鲁棒性,控制结果显示系统具有良好的静动态性能.

Aimed at the problems of time varying parameters and large load disturbance in cylinders servo system of hydraulic-driven parallel robot, a new dynamic neural network was designed by means of adding feedback loops to feedforward network, and the intelligent controller based on this dynamic network was employed to control the cylinder servo system. The network learning algorithm was deduced according to the new integrative objective function that evaluated system performance perfectly. The experimental results show that the controller design is independent upon system model, and the controller has self-adaptability to time varying parameters and robustness to large load disturbance with the satisfactory static and dynamic performance.