论文部分内容阅读
本文提出了一种设计机器人模糊－神经控制器的新方法。因为模糊逻辑控制器的控制信号是由系统的响应行为而不是由其分析模型决定的，所以机器人的开环响应可以用一系列二阶系统来描述。然后，用Ｎｅｌｄｅｒ－Ｍｅａｄ单纯形算法离线优化与该系统相关的模糊逻辑控制器的参数，并用人工神经网络来训练开环响应与这些优化了的参数之间的匹配关系。该方法的优点在于当它用于设计模糊－神经控制器时，在自适应过程中不必考虑收敛问题
This paper presents a new method of designing a robot fuzzy-neural controller. Because the fuzzy logic controller’s control signal is determined by the system’s response rather than by its analytical model, the robot’s open-loop response can be described by a series of second-order systems. Then, the Nelder-Mead simplex algorithm is used to off-line optimize the parameters of the fuzzy logic controller related to the system. The artificial neural network is used to train the matching relationship between the open-loop response and the optimized parameters. The advantage of this method is that when it is used to design a fuzzy-neural controller, there is no need to consider the convergence problem in the adaptation process