An Auto-Tuning Controller Using Recursive Fuzzy Reasoning

Fuzzy reasoning is able to utilize human experimental rules for driving numerical results. So, recently, some automatic control systems based on fuzzy reasoning have been developed. But the conventional fuzzy reasoning which is used in control has difficulties for application to parameter estimation. Because it needs inputs which are always meaningful and doesn't have property like convergence. In this paper, we propose recursive fuzzy reasoning. This reasoning inputs phenomena induced by a parameter and estimates the value of the parameter, gradually. This algorithm has a learning membership function which narrows range of the parameter. We have applied the recursive fuzzy reasoning to a PI autotuning controller. Usual controlled systems are so complicated that a skilled human operator's technique is often more desiable than that based on mathematical models. Besides, the auto-tuning is a kind of parameter estimation. So the recursive fuzzy reasoning is suitable for this problem. Furthermore, we use other two new techniques for this application. They are the evaluation of unsatisfaction and the extraction of characteristic parameters. Simulation tests are presented. The recursive fuzzy reasoning automatically tunes control parameters during ordinaly on-line control operation without any special test signal.