论文部分内容阅读
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。
Robots are strongly coupled nonlinear dynamical systems. In order to design its controller, a dynamic neural fuzzy system is used to study the nonlinear H∞ control problem. Based on the traditional T-S neural fuzzy system, delay feedback and memory unit are introduced into it, and the corresponding dynamic BP training algorithm is derived for this network. This network can effectively reflect the dynamic performance of nonlinear systems such as robots and overcome the shortcoming that static neural network can not effectively represent the dynamic system mapping relationship. Based on this fuzzy model, the design of system controller is studied by H∞ method. Finally, the simulation of inverted pendulum shows that this controller has good robustness.