全局稳定的机器人CMAC标称补偿跟踪控制研究

GLOBALLY STABLE TRACKING CONTROL OF ROBOT MANIPULATORS BASED ON CMAC DESIRED COMPENSATION

提出了一种新的小脑模(Cerebellar Model Articulation Controller,CMAC)神经网络标称补偿控制器.采用二阶扩展B样条CMAC网络平滑逼近机器人标称模型,消除了常规神经网络控制对输入的严格假设.为了确保系统闭环的全局稳定性,采用Lyapunov直接法设计网络权值的更新律,并引入非线性反馈项完全抵消补偿的残留项.未知的CMAC逼近误差和系统随机干扰,通过一个简洁的鲁棒自适应律估计.最后,针对两自由度机器人的仿真实例验证了所提算法的有效性.

A new neuro-controller based on the Cerebellar Model Articulation Controller (CMAC) desired compensation is proposed in this paper. A 2nd-order expended B-spline CMAC is used to approach the desired model of robot manipulators and it eliminates the rigid assumption on Neural Network (NN) inputs which is needed in conventional neuro-controller. In order to insure the global stability of the closed-loop systems, the weights update law is designed based on the Lyapunov direct method, and the nonlinear feedback terms is introduced to fully counteract the compensated residual terms. A simple robust adaptive law is used to estimate the unknown approach errors and bound random disturbances. Finally, the proposed algorithm is verified through simulation study on a two degree of freedom (2-DOF) robot.