一种新的CMAC实时在线控制器

A new CMAC neural network's online control

针对两关节焊接机械手的自适应实时控制,提出了一种双CMAC网络加PD控制器的实时自学习控制器结构。利用两个信度指派CMAC网络分别逼近两关节机械手的不同关节的逆模型,大大减少了CMAC网络需要的存储空间,避免了必须的HASH地址映射;在CMAC加PD控制器结构中采用了不同于传统的CMAC网络权值修正方法,不仅提高了模型的逼近精度,同时将CMAC运行所需要的时间减少为原来的一半,增强了PD参数的鲁棒性,更加有利于实时控制的进行。

With the two -link manipulator＇s online control, a two CMAC NN plus two PD controllers structure is introduced. In this structure, two credits assigned CMAC NN is used to approach two manipulator joint＇s inverse model. The memory of CMAC need is reduced a lot and the hash mapped is left out. Compared with the traditional controller, the new controller structure is more precise, and the time for one step is the half of that,It improved the robust of PD parameters and the performance of the real time controller. The track of the two -link manipulator shows that this new controller structure has more robust ability and is feasible.