Automated Mission Planning via Evolutionary Algorithms

Abstract

Many space mission planning problems may be formulated as hybrid optimal control problems, that is, problems that include both real-valued variables and categorical variables. In orbital mechanics problems, the categorical variables will typically specify the sequence of events that qualitatively describe the trajectory or mission, and the real-valued variableswill represent the launchdate,flight times betweenplanets,magnitudes anddirections of rocket burns, flyby altitudes, etc. A current practice is to preprune the categorical state space to limit the number of possible missions to a number whose cost may reasonably be evaluated. Of course, this risks pruning away the optimal solution. Themethod to be developed here avoids the need for prepruning by incorporating a new solution approach. The new approach uses nested loops: an outer-loop problem solver that handles the finite dynamics and finds a solution sequence in terms of the categorical variables, and an inner-loop problem solver that finds the optimal trajectory for a given sequence A binary genetic algorithm is used to solve the outer-loop problem, and a cooperative algorithm based on particle swarm optimization and differential evolution is used to solve the inner-loop problem. Thehybrid optimal control solver is successfully demonstrated here by reproducing theGalileo andCassinimissions.