LQR Discrete Controller Tuning for a TWIP Robot Based on Genetic Algorithms

Abstract

A two wheeled inverted pendulum (TWIP), is an unstable system with highly non-linear dynamics. Therefore, it is necessary to design controllers that can react properly to different kinds of external disturbances, keeping tracking accuracy and robustness. The aim of this paper is tuning a Linear Quadratic Regulator (LQR) discrete controller using genetic algorithms (GA) for improve the reference tracking performance. The GA consider the ISE index in the reference tracking for the fitness function in order to obtain the optimal LQR controller parameters Q and R and state-feedback gains. Finally, test simulations are performed to compare the behavior of traditional LQR and the extended LQR-GA controller.