Auto-tuning controllers of a class of plants using gradient descent algorithm

Nguyen Trong Toan, Nguyen Nam Trung

In industry, PID controller is widely used in the regulator of technology processes.  Many methods of designing conventional PID controllers do not work well for time-delayed linear, vague, and nonlinear systems. Therefore, PID parameter adjustment plays an important role. To solve this problem, many recent studies have focused on the development of auto-tuning controllers. Using the method of gradient descent algorithm to the parameter space of the controller whose coefficients developed by adaptive algorithms in this paper is adjusted automatically online to minimize the performance index without knowing the exact mathematical model of the plant. The simulation and test results on real plant are presented to illustrate the performance of the proposed auto-tuning PID controllers and a comparison with its traditional PID controller is made.