Massively parallelized co-evaluation for many-objective space trajectory optimization
Martin Schlueter, Masaharu Munetomo · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2018
This contribution presents numerical results for optimizing a many-objective space mission trajectory benchmark under consideration of massively parallelized co-evaluation of solution candidates. The considered benchmark is the well-known Cassini1 instance published by the European Space Agency (ESA) extended to four objectives. The MIDACO optimization software represents an evolutionary algorithm based on Ant Colony Optimization (ACO) and is applied to solve this benchmark with a varying fine-grained parallelization factor (P) ranging from one to 1024. It can be shown that the required number of sequential steps to solve this benchmark can be significantly reduced by applying massive parallelization, while still maintaining a sufficient high number of well distributed non-dominated solutions in the objective space.