Integrating PSO Optimized LQR Controller with Virtual Sensor for Quadrotor Position Control

Abstract

Linear Quadratic Regulator is one of robust and optimal controller that mostly used for handling Multiple Input Multiple Output (MIMO) system. Although, an LQR controller can handle MIMO system, it is difficult to determine the optimal weighting matrices to achieve optimal performance. An alternative for optimizing these matrices is by introducing Particle Swarm Optimization (PSO) method. Furthermore, not all state variables of a system to be controlled are available for measurement due to lack of reliable sensors, which leads to the development of virtual sensing technology. This is another alternative in control application since it can replace actual real sensors with software approximation. In this paper, development of a PSO optimized LQR controller integrated with virtual sensing system will be introduced. The developed virtual sensor consists of a Diagonal Recurrent Neural Network (DRNN) and coupled with Extended Kalman Filter (EKF), which can estimate the unknown variables from the a priori known variables. The designed control strategy will be tested on a quadrotor model having 12 states variables. The simulation results show how the position of the quadrotor can be controlled optimally and satisfactorily. Comparison with PID based controller also confirms the superiority of the proposed control system.