Improving a linear–quadratic regulator controller by genetic algorithm on a Quanser gyroscope

Abstract

In this paper, we present the experimental evaluation of an optimal control solution for the Quanser gyroscope system. To address the challenge of manually tuning the Q and R weighting matrices of the traditional Linear–Quadratic Regulator (LQR) controllers, which performs poorly when system parameters vary, we developed an optimal controller based on a Genetic Algorithm. This method automatically searches for optimal Q and R values based on the Integral of Absolute Error (IAE) criterion, ensuring precise trajectory tracking and fast response. Experimental results demonstrate that the proposed optimal controller exhibits superior performance, with a settling time of 1.98 s, zero overshoot, and a steady-state error of only 0.21 degrees. The most significant contribution of this study, however, is the rigorous experimental implementation and validation of the actual Quanser hardware. Data from the hardware demonstrates the controller's superior, stable operation, enhanced accuracy, and robust dynamic response compared to standard LQR in a real-world environment. These results affirm the potential of the developed method, even when the system is subjected to noise, delays, and nonlinearities. This paper highlights the crucial role of hardware validation in translating theoretical advancements into reliable and effective practical control solutions.