Adaptive robust neural fuzzy control of uncertain systems: A Lyapunov theory approach

The objective of this research was to develop effective control strategies for uncertain nonlinear dynamical systems. In the first stage of the research, neural fuzzy controllers were proposed. Genetic algorithms were employed to design and fine-tune the proposed neural fuzzy controllers, which then were tested on an anti-lock brake system model and a ground vehicle. Training or fine-tuning of the above described controllers was performed off-line and found to be time consuming. To overcome this problem, an adaptive control algorithm was developed that learns and compensates for the unmodeled dynamics of the plant online. In addition, a robustifying component was proposed whose role is to suppress modeling errors and uncertainties. Integrating the adaptive and the robust approaches resulted in a guaranteed transient tracking performance and a guaranteed final tracking error accuracy in the presence of modeling errors and disturbances. The closed-loop system driven by the proposed controllers was shown, using the Lyapunov method, to be stable with all the adaptation parameters being bounded. To further enhance the performance of the proposed control strategies, a self-organizing raised-cosine radial basis functions (RCRBFs) component was included in the control architecture. The proposed self-organizing RCRBF network can adjust its size by growing or shrinking the number of basis functions used according to the design specification. Performance comparison of the proposed controllers with the one in the literature was conducted. The proposed controller outperformed the controllers found in the literature