Robust control system design using simulated annealing

Design parameters of a flight control system are optimized by a probabilistic method. Simulated annealing is applied for the optimization, and the downhill-simplex method is added to generate new design vector candidates. The cost function to be minimized is chosen as the probability of violating the design criteria, and it is derived by Monte Carlo evaluation that incorporates various uncertainties. Thus, the designed system is robust against these uncertainties. The feasibility of the algorithm is demonstrated by designing a control system for a simplified model. The results show that simulated annealing is more effective than the downhill-simplex method for parameter optimization, and it requires less computational time than the genetic algorithm. The Automatic Landing Flight Experiment unpiloted reentry vehicle provides a second example. Simulated annealing is shown to produce a more robust longitudinal flight control design than that used in the 1996 flight experiment.