Stable hybrid control based on discrete-event automata and receding-horizon neural regulators (Q5943846)

A hybrid control scheme for nonlinear discrete-time systems is developed. First, a finite set of receding-horizon feedback control laws based on neural approximators is determined off-line. Second, a discrete-time discrete-event automata is introduced which is aimed at choosing the best control law from a finite set depending on-line on the current conditions of the system and on external events. The stability analysis of the proposed hybrid control system is carried out. The stabilizing properties of individual control laws, restrictive conditions in the switching time instants, tuning algorithms guaranteeing uniform bounds to the approximation error, and other aspects of hybrid control are discussed. The effectiveness of the proposed control scheme and its practical applicability is demonstrated on a complete simulation analysis of a real-world freeway traffic control problem.