Implement the Fuzzy Controller by Imitating the Tuned PID Controller Using Reinforcement Learning

Abstract

This article presents how to design a fuzzy controller for controlling engine speed. The fuzzy controller is designed to mimic the effect of the PID controller, where the PID controller is adjusted using reinforcement learning. Controller performance is compared between a controller tuned with control system tuning and one tuned using reinforcement learning. While PID tuning with the control system is fast and effective for model-based systems, reinforcement learning tuning is suitable for highly nonlinear systems. The result of the experiment is to use the reinforcement learning agent to calculate the gain of PID, after which the fuzzy logic learned in tuned PID is generated. Fuzzy logic control is easy to adjust for nonlinear systems control. Fuzzy inference system tuning both using ANFIS and pattern search are compared.