Novel approach for knowledge-based evolutionary parameter tuning for controllers

There are three types of parameter tuning for controllers: start-up, on-demand and evolutionary. This paper proposes a novel approach of knowledge-based evolutionary parameter tuning for controllers. First, a knowledge-based system is developed for online pattern recognition of transient responses of feedback control systems. Then, iteration of controller parameters is formulated as a quasi-LQR (linear quadratic regulator) problem. A set of rules are designed to supervise the embedded utilization of solution to the quasi-LQR. Effectiveness of the proposed approach is demonstrated by simulations