Sequence optimization for multiple asteroids rendezvous via cluster analysis and probability-based beam search

Abstract

It is of great significance to visit multiple asteroids in a space mission. In this paper, the multiple asteroids mission optimization is implemented using cluster analysis and probability-based beam search. Clustering is performed to select the first asteroid to visit. Four cluster algorithms are investigated and affinity propagation is selected. Then four beam search algorithms that are deterministic beam search and three probability-based beam search variants, probabilistic beam search, ant-colony beam search, and evolving beam search, are applied to search for the rendezvous sequence. Deterministic beam search as a heuristic tree search algorithm is widely applied in multitarget sequence optimization, but it has an obvious drawback of the conflict between the number of pruned nodes and the possibility of finding optimal solutions, which can be improved by probability-based beam search. Among three probability-based beam search, the ant-colony beam search has a learning mechanism, and evolving beam search is constructed based on ant-colony beam search and has an evolutionary mechanism. Results show that the introduction of randomness can improve beam search, and beam search variants with the learning and evolutionary mechanism have an excellent performance.