Evolutionary optimization of interplanetary trajectories: improvements from initial diversification

Abstract

A diversification phase based on tabu search is introduced to improve the performance of an evolutionary algorithm, which runs a genetic algorithm, differential evolution, and particle swarm optimization in parallel, according to the island model. Two problems concerning space trajectory optimization (Cassini mission and a round trip mission to near-Earth asteroids) are considered. These problems present different peculiarities which make the search of the global optimum difficult. It is shown that the probability of success may be, in some cases, quite low, due to the presence of many local optima and/or the existence of a bias towards suboptimal solution. The addition of an initial diversification phase in most cases improves the algorithm capability of finding the global optimum, without significantly increasing (and sometimes even decreasing) the number of function evaluations required to attain the optimum and thus reducing the required computational effort.