This research successfully develops a new numerical method for the problem of two-sided flight path optimization, that is, a method capable of finding trajectories satisfying the necessary condition of an open-loop representation of a saddle-point trajectory. The method of direct collocation with nonlinear programming is extended to find the solution of a zerosum two-person differential game by incorporating the analytical optimality condition for one player into the system equations. The new method is named semi-direct collocation with nonlinear programming (semi-DCNLP). We apply the new method to a variety of problems of increasing complexity; the dolichobrachistochrone, a problem of ballistic interception, the homicidal chauffeur problem and minimum-time spacecraft interception for optimally evasive target, and thus verify that the method is capable of identifying saddle-point trajectories. While the method is quite robust, ambitious problems require a reasonable initial guess of the discretized solution from which the optimizer may converge. A method for generating a good initial guess, requiring no a priori information about the solution, is developed using genetic algorithms. The semi-DCNLP, in combination with the genetic algorithm-based preprocessor, is then used to solve a very complicated pursuit-evasion problem; optimal air combat for realistic fighter aircraft models in three dimensions. Characteristics of the optimal air combat maneuvers for both aircraft are identified for many different initial conditions.
展开▼