Multiobjective Design Exploration of a Many-objective Space Trajectory Problem for Low-Thrust Spacecraft Using MOEA with Large Populations

摘要

This study explores many-objective trajectory designs for the future low-thrust spacecraft proposed in ISAS/JAXA. The various trajectory profiles are identified by multiobjective evolutionary algorithm (MOEA) for many-objective optimization. One difficulty of the low-thrust transfer problem is that there are many possible trajectory profiles due to many revolutions during the orbit raising phase. The objective functions of the low-thrust transfer problem are designed to minimize (1) the operation time of the Ion Engine System, (2) the flight time to reach the lunar orbit, (3) the maximum eclipse time and to maximize (4) initial mass of the spacecraft. For improving the performance of the evolutionary computation with large populations, CHEbyshev-Epsilon opTimizer AlgoritHm (CHEETAH) is adopted. The CHEETAH combines e-indicator and Chebyshev achievement function into a ranking method, and is designed for the parallel evaluation. The parallel CHEETAH with large populations is conducted at "K" supercomputer. The analysis of nondominated solutions reveals various trade-off relations and correlations among the objective functions. Furthermore, the analysis results provide useful knowledge to design trajectory profiles in the low-thrust spacecraft with the ion engine.