In this paper, a novel learning-based fuzzy Linear Quadratic Regulator (LQR) control method using Extended Kalman Filter (EKF) to optimize a Mamdani fuzzy LQR controller is presented. The EKF is used to adjust the shape of membership functions and rules of the fuzzy controller to adapt with the working conditions automatically during the operation process to minimize the control error. Then, the LQR controller is tuned according to the modified fuzzy membership functions and rules. The proposed approach in this paper is verified by testing and comparing performance of the height control problem of an unmanned quadrotor helicopter between the conventional LQR and learning-based fuzzy LQR controllers in the Matlab/Simulink. Simulation results show that developed method is effective for online optimization of fuzzy LQR controllers, improving control performance significantly.
展开▼
