Lyapunov stability analysis for self-learning neural model with application to semi-active suspension control system

A supervised training algorithm and an unsupervised learning algorithm for neural models are derived using Lyapunov stability theory and the concept of model reference adaptive control (MRAC). The practicability of the algorithm is demonstrated by its successful application to a semiactive suspension system. The neurocontroller establishes its own control laws for improving the suspension performance without requiring a complete knowledge of the system dynamics. It also makes it possible to evaluate the roles of simple sensors in defining the control laws. Simulation results illustrate the self-improving suspension performance and show that superior suspension performance can be achieved using the neurocontrol. It is found that the discretized version of the developed training and learning algorithms corresponds to certain well-known existing algorithms