On closed-loop controls in pursuit-evasion

Abstract

A hybrid algorithm is introduced in order to generate optimal closed-loop controls in a pursuit-evasion problem. The algorithm is composed of a differential dynamic programming method and a singular perturbation technique. The first mentioned method is used in those parts where the latter is weak. The method so created has an open-loop character. By having the open-loop solution updated as often as possible when new measures of the states have been estimated, the method will be near-optimal closed-loop. The game problem is discussed in terms of optimistic or pessimistic strategies, based on assumptions of the evader's strategy. In practice, the evader he may not do his best. The demonstrative example is in the horizontal plane, formulated as a game of degree. The pursuer uses a variable speed model with throttle and turn controls. The computer requirements can be kept to a reasonable level, and the new algorithm can be implemented on an airborne computer.