On the Design of Efficient Global Search Algorithms for Spacecraft Trajectory Optimization Problems

Abstract

Global optimization of spacecraft trajectories is often a time-consuming process because optimal spacecraft trajectory problems tend to be high-dimensional with complex feasible domains. This thesis aims to contribute new models and algorithms for the development of more efficient global search processes as they apply to spacecraft trajectory optimization problems. Three low-dimensional trajectory optimization problems in order of increasing complexity are designed to test out the novel methods used in the rest of the research process. Methods to uncover the feasible solution space are developed and applied to the aforementioned trajectory optimization problems. The solution spaces are visualized and analyzed to identify local minima, and a clustering algorithm is used to approximate the basins of feasible solutions surrounding these local minima based on generated data. Finally, an adaptive sampling algorithm informed by the structure of the solution space is developed and compared to a naive uniform sampling algorithm. The adaptive sampling algorithm is shown to uncover a larger collection of qualitatively different solutions in a fixed time, which is an important step in finding the global optima.