Hybrid self-learning fuzzy PD+I control of unknown linear and nonlinear systems

A human being is capable of learning how to control many complex systems without knowing the mathematical model behind such systems, so there must exist some way to imitate that behavior with a machine. A novel hybrid self-learning controller is proposed that is capable of learning how to control unknown linear and nonlinear processes incorporating a human-like learning behavior. The controller is comprised of a Fuzzy PD controller plus a conventional I controller and its corresponding gains are tuned using a human-like learning algorithm in order to reach specified goals of steady-state error (SSE), settling time (Ts) and percentage of overshooting (PO). Among the systems tested are first and second order linear systems, nonlinear pendulum and the nonlinear equations of Van der Pol, Rayleigh and Damped Mathieu. Analysis and simulation of a second order linear and nonlinear pendulum is provided to demonstrate that the proposed controller has excellent results.