Diversity-based heuristic search for multiple-asteroid tours
Date published
Free to read from
Supervisor/s
Journal Title
Journal ISSN
Volume Title
Publisher
Department
Type
ISSN
Format
Citation
Abstract
An asteroid tour is a sequence of asteroid flybys performed by a spacecraft in a single mission. Visiting different types of asteroids would be of a great scientific return, but planning such type of mission is highly complex. This is because the combinatorial problem of selecting sequences of asteroids is coupled with optimal control problem of finding viable trajectories in terms of transfer times and manoeuvre locations. The combinatorial problem can be tackled with heuristic algorithms that explore the search space to find optimized flyby sequences of asteroids. This paper proposes a search strategy to increase the diversity of the set of flyby sequences found from the resolution of the optimization problem. The goal is to find a tree exploration method that will improve the quality of the solutions by increasing the number of targets encountered and eliminating duplicated sequences while preserving the average ∆v cost. A diversity measure for a set of solutions is defined based on the ∆v cost of the transfer and the nature of the asteroids in the sequences. A variant of the Beam Search (BS) algorithm, referred to as Diversity Search (DS), is tested. Unlike BS, the DS algorithm eliminates duplicated sequences from the final set of solutions. New asteroid tours are found, allowing to reach targets with a higher inclination or eccentricity, but the average ∆v of a flyby sequence is higher than solutions found with BS. A hybrid search strategy using BS and DS is explored. The goal is to reduce the search space with respect to two criteria: ∆v and diversity score. By pruning the search space according to ∆v first, the hybrid search algorithm is able to generate a diverse set of asteroid tours while preserving the total ∆v of the solutions. Compared to DS, less new asteroids are encountered in the solutions but the average ∆v remains similar to results obtained with BS.