Cell state space based incremental best estimate directed search algorithm for Takagi-Sugeno type fuzzy logic controller automatic optimization

This paper presents a new cell state space based Takagi-Sugeno type fuzzy logic controller (FLC) automatic optimization algorithm, called an incremental best estimate directed search. Starting from an initial FLC with poor performance, the corresponding control surface is sampled over all the controllable cell centers, the sampled data set is called the training data set. Using the least mean square learning algorithm with the training set, another FLC with randomly initialized parameters is trained in an iterative procedure. In each iteration, the trained FLC is evaluated with cell mapping based global and local performance measures, the training set is then updated based on the evaluation with best kept policy. In this way, the training set is optimized in every iteration, and the FLC trained by the training set is also optimized progressively. Since the new algorithm makes use of every FLC evaluated, a fast convergence speed is expected. A 4D inverted pendulum is studied to justify the methodology.