挖掘机器人伺服系统神经网络滑模控制

挖掘机器人伺服系统存在高度非线性、参数不确定和未建模动态等诸多不利因素,提出了一种结合径向基函数(RBF)神经网络的非线性滑模控制器,以提高控制精度和鲁棒性。首先,建立了单联伺服系统的数学模型;其次,采用RBF神经网络对系统的不利因素进行逼近,提出积分滑模面进一步减小稳态误差,同时减少对伺服系统参数的依赖,在此基础上,设计了基于RBF神经网络的滑模控制器(SMC-RBF),利用Lyapunov理论证明了系统的渐近稳定性;最后,通过不同的参考信号和整平实验验证了控制器的优越性。仿真结果表明,SMC-RBF控制器响应快,跟踪精度高且鲁棒性强,与PID控制器相比正弦轨迹跟踪精度提高了46%。整平实验结果表明,铲斗末端轨迹跟踪精度提高了52%。

Neural Network Slide Mode Control of Robotic Excavator Servo Systems

Controlling the servo system of a robotic excavator is very difficult since it faces with a lot of unfavorable factors such as high nonlinearity, parameters uncertainty and unmodeled dynamics. A nonlinear integral sliding mode controller combined with radial basis function (RBF) neural networks is presented to improve the servo system accuracy and robustness. Parameters perturbation and uncertain nonlinearity are compensated by RBF neural networks, the integral sliding surface can reduce the steady-state error. Model of the bucket servo system is established firstly. On this basis, the sliding mode controller combined with RBF neural networks (SMC-RBF) is designed. Furthermore, the asymptotic stability of the designed control method is certificated by Lyapunov theory. Finally, different reference signals and experiment are introduced to test the superiority of the presented control method. Simulation results indicate that the SMC-RBF control method not only meet requirements of high precision tracking performances, but also show some robustness. Sinusoidal trajectory tracking precision of the SMC-RBF control method is improved by 46% compared with PID controller. SMC-RBF control method can improve the trajectory tracking precision of a two meters straight line by 52%.