A modified Smith predictor for the control of systems with time–varying delay

A conventional Smith predictor presents poor stability when controlling systems with time-varying delay. In this paper, an improved adaptive PID-Smith predictor is proposed. It uses a PID controller as the primary controller as well as the estimator for unknown time delay. The goal is to ensure system stability and resistance to modeling errors.
This article discusses two structures of the estimator unit - based on a neural network and on a fuzzy controller. In the first variant, the genetic algorithm is used to find the optimal parameters of the estimator in the autonomous mode. In the second variant, the fuzzy controller of the Takagi – Sugeno type uses a variety of models with different delay time. At each time point the error of output is calculated for all models. The output signal of the estimator is formed by the rule of defuzzification. Simulation results show the effectiveness of the proposed modification of the Smith predictor.