GA&LS-based Fuzzy Neural Network control

With its stochastic but directional research mechanism, Genetic Algorithm (GA) was used to optimize the global parameters and structure of Fuzzy Neural Network (FNN), membership functions of premise of network reasoning rule were adjusted by GA offline, while network weights of outcome were dynamically modified by least square (LS) algorithm online. The two algorithms worked harmoniously so as to improve the performance of network. Rotary inverted pendulum was utilized to examine control performance and feature of the approach proposed. The result of experiment shows the validity of method presented, the system has good self-learning capability and strong robustness.