A hybrid genetic algorithm approach to global low-thrust trajectory optimization

Matthew A Vavrina, Purdue University

Abstract

To expand mission capabilities that are required for exploration of the solar system, methodologies to design optimal low-thrust trajectories must be developed. However, low-thrust, multiple gravity-assist trajectories pose significant optimization challenges because of their expansive, multimodal design space. In this work, a technique is developed for global, low-thrust, interplanetary trajectory optimization through the hybridization of a genetic algorithm and a gradient-based direct method (GALLOP). The hybrid algorithm combines the effective global search capabilities of a genetic algorithm with the robust convergence and constraint handling of the local, calculus-based direct method. The automated approach alleviates the difficulty and biases of initial guess generation and provides near globally-optimal solutions. Both single objective and multiobjective implementations are developed. In the single objective implementation, the technique is applied to several complex low-thrust, gravity-assist trajectory scenarios, generating previously unpublished optimums. Specifically, the single objective hybrid algorithm generates apparent global optimums for a direct trajectory scenario to Mars, as well as gravity-assist trajectories with three intermediate flybys to Neptune and Pluto. The multiobjective implementation incorporates the NSGA-II algorithm to generate a Pareto front of solutions that are globally optimal in terms of both final delivered mass and time-of-flight in a single execution. The multiobjective hybrid algorithm is applied to a direct Earth-Mars rendezvous design scenario, successfully developing a Pareto front of near-globally optimal trajectories, enabling a tradeoff decision on the two objectives.

Degree

M.S.

Advisors

Howell, Purdue University.

Subject Area

Aerospace engineering

COinS