Study: parametric performance of island model optimisation applied to interplanetary trajectories

The optimisation of interplanetary trajectories is a challenging field. Nowadays many optimisation techniques rely on the islands model, but few analysis about the performance of such technique have been carried out. The goals of this study are:- Analysis of the performance of the islands model optimisation algorithm applied to several interplanetary trajectories scenarios.- Parametric optimisation of the islands model configuration parameters.
The study of interplanetary trajectories is a field with many difficulties due to the complexity of the problem at hand. Finding a trajectory that fulfils all the position, time and fuel consumption requirements is a complicated task. In order to find good trajectories, this study uses a multi-gravity assist based approach, coupled with an island model optimiser that executes in parallel some variants of genetic algorithm, differential evolution and particle swarm optimisers. On this thesis, the algorithm will be applied to three different scenarios: a fictional Earth-Mars direct launch and the Mariner 10 and Voyager 2 missions. These have been selected to represent a broad range of scenarios: a direct launch between two planets, a mission towards the inner planets of the Solar System and a mission directed towards the outer ones. The first goal of the study is to determine whether using an island model improves the performance of the genetic algorithm, differential evolution and particle swarm algorithms executed stand-alone (without the island model). Then, we analyse whether mixing algorithms (heterogeneous island model) provides any benefits to the simple island model. Ultimately, a parametric performance analysis is carried out to determine the behaviour of the island model parameters. The results show that using an island model in an algorithm improves its performance in most cases. Both methods find the same optimal regions, but the island model offers more accurate results. Additionally, mixing algorithms in the island model helps achieving a higher stability, but, at the same time, it lowers the probability of finding the best solution. The island model, in a general fashion, shows an improvement when the number of islands is equal or inferior to 10. It also usually needs a minimum of 20 global generations to find the best solution, while the heterogeneous island model shows better results with less than five global generations. The use of the migration ope