A novel method of self-tuning PID control system based on time-averaged Kalman filter gain

This paper presents a study on the use of Kalman filter in estimating the PID controller parameters, which are assumed to have the same numerical values for either proportional, integral or derivative modes. The plant is forcedly treated as a linear first order, and only rough information about it is provided. The process and output measurement are disturbed with white noise. Apart from the PID controller, a one-dimensional Kalman filter is added to the closed loop system in order to estimate the controller parameters, represented by Kalman gain. Since the actual system is in general more complicated than that of the first order, its parameters must be tuned (estimated) every sampling process, and consequently they are time varying and function of the system state. From the simulation it is shown that the three parameters of PID controller can be best represented by the time-averaged Kalman filter gain obtained during the sampling of the process.