Application of soft computing techniques to a LQG controller design

Abstract

Optimal control with a linear quadratic Gaussian (LQG) controller is a very popular and a modern control methodology. However, the optimization of design matrices of a linear quadratic regulator (LQR) and Kalman filter is a time consuming process and needs a significant amount of effort. Herein, soft computing techniques are proposed to automate this process. The noise covariance matrix V is made adaptive by the use of fuzzy logic. The Kalman filter is calculated each time with a new value of noise covariance matrix. The fuzzy Kalman filter is then combined with the LQR to form a novel LQG controller. This approach is highly desirable as it eliminates the painstaking heuristic procedure and improves the quality of the Kalman filter and hence the performance of LQG. This approach has been implemented on a twin Rotor MIMO system (TRMS) which is similar to a helicopter. This paper discusses the TRMS modelling using system identification techniques, fuzzy LQG controller design and implementation using the data collected from the TRMS.