LQR controller tuning by using particle swarm optimization

Abstract

LQR is an optimal controller. Optimal in that it is defined so as to provide the smallest possible error to its input. Q and R matrix of LQR usually selected by trial and error. In two wheeled inverted pendulum robot, the most important variable to control is tilt angle. Therefore in this thesis, the value of Q is firstly set and then R the identity matrix is set. For small rising time and low overshoot for the overall control. After getting good value of Q, the feedback gain K is obtained. By using MATLAB simulink, we simulated new PSO algorithm for the LQR control to select the best Q control matrix. The selection is based on the smallest integral of absolute error of the random Q. From the simulation results, the very challenging controller design for the TWIP control system has been realized by the PSO-based LQ regulator. It is our firm belief that the proposed method is use useful not only for the control of TWIP robot problem but also for other difficult problems.