A Comparison of CMAC Neural Network and Traditional Adaptive Control Systems

Abstract

A neural network based controller similar to Miller's CMAC method [6] is compared to a self-tuning regulator [2] and a Liapunov based model reference controller [8]. The three control algorithms are tested on exactly the same control problems. Results are obtained for the case where the system being controlled is linear and noise free, for the case where noise is added to the measurements, and for the case where a non-linear system is controlled. Comparisons made with respect to closed-loop system stability, speed of adaptation, noise rejection, robustness, the number of required calculations and system tracking performance indicate that the neural network approach exhibits the potential to solve some of the problems that have plagued more traditional adaptive control systems.