机械手分散CMAC控制器

针对机械手动力学模型未知情况 ,采用类似于计算力矩法的控制器结构 ,设计了基于 CMA C神经网络分散控制器作为非线性估计器、PD控制器作为跟踪外环的机械手控制方案。分析了该控制方案的稳定性 ,通过仿真验证了该控制方法的有效性。结果表明 ,在未知机械手动力学的情况下 ,控制系统的运行仍是有效的 ,并且不同于传统的自适应控制器 ,不需要计算回归矢量 ,也不必对系统参数做任何线性假设。由于采用分散CMAC结构 ,简化了控制器实现的复杂度 ,易于采用大规模可编程数字逻辑阵列和 ASIC数字电路予以硬件实现。

Decentralized CMAC Neural Network Control of Robot Manipulators

A control scheme of robot manipulator which has CMAC based neural network decentralized controller as inner-loop and a PD contr oller as outer-loop is presented for a class of robot manipulators with unknown dynamic model. The stability is analyzed, and its validity is also demonstrated by numerical simulation. Under the unknown dynamics of the robot manipulator, t he controller is still valid. In contrast to traditional ones, the control schem e does not require the computation of a regression vector and need not make any linearity assumptions on the system parameters. Because of the decentralization ,a large-scale logic array and ASIC digital circuit can realize the CMAC contro ller easily.