机械手动态T-S神经模糊H∞控制器设计

Dynamic neuro-fuzzy H∞ controller design of robotic manipulators

机器人是强耦合的非线性动力学系统。为了设计其控制器,利用动态神经模糊系统对非线性H∞控制问题进行了研究。在传统的T-S神经模糊系统基础上,将延迟反馈和记忆单元引入其中,并针对此网络推导了相应的动态BP训练算法。利用此网络可以有效反映机器人等非线性系统的动态性能,克服了静态神经网络无法有效表示动态系统映射关系的缺点。在此模糊模型的基础上,采用H∞方法研究了系统控制器的设计。最后以倒立摆为例的仿真试验表明此控制器具有良好的鲁棒性。

A dynamic neuro-fuzzy H~∞ control approach was developed for trajectory tracking of a robotic manipulator with poorly known dynamic nonlinearities. The method is based on a dynamic T-S neural-fuzzy network with feedback delay and memory cells. A dynamic back propagation (BP) learning algorithm was used to train the system. This type of network can describe the mapping relationship of non-linear dynamic systems such as robots, which can not be effectively characterized by static networks. The movement of an inverse pendulum was tracked to demonstrate the good robust performance of the controller design.