Control law design using repeated trials

The paper presents an adaptive random search algorithm (ARS) for control system design. The algorithm is similar to L. Pronzato et al.´s (1984) ARS algorithm except for the use of a probability of acceptance function that determines the likelihood of choosing a candidate set for trial based on an estimate of the likely success of that set. The function is parameterized by a set of adaptively chosen interpolation points and function interpolation rules. The algorithm is first used to design the longitudinal control laws for an unmanned air vehicle (UAV). Then it is used to design the lateral/directional control laws for the UAV. Simulation results show that: a) this technique can be used for a practical control system design; b) the time to converge to an acceptable solution in a real system can be decreased significantly by the use of the adaptively adjusted probability function that is used to guide the random search