Tuning a PID controller using genetic algorithms

Abstract

This paper details the development of PID controller tuning, based on the implementation of advanced optimization techniques in  MATLAB to find the optimal gains for control actions. The methodology used to create the solutions was that of genetic algorithms, an artificial intelligence technique developed in the 1970s and inspired by Darwin's natural selection, within the field of evolutionary computing. Its implementation is based on selection, crossover, and mutation processes, which allow the solutions to iteratively converge towards increasingly optimal results. Two different genetic algorithms were programmed and designed. The first focused exclusively on a single objective, which was the settling time; while the second was based on a multi-objective technique that additionally considered the maximum overshoot, rise time, and delay time. Different fitness functions were developed to create these neural models; subsequently, the gain results obtained from these genetic methods were compared with those proposed by analytical and experimental methods, both in the field of simulation and in physical implementation. The analysis of the responses validated the
efficiency and effectiveness of the proposed algorithms for controller tuning, showing better performance with the gains obtained through genetic algorithms.