Multi-objective global optimization for interplanetary space trajectory design

This contribution presents numerical results for global optimization of a multi-objective formulation of the well-known Cassini1 interplanetary space trajectory benchmark published by the European Space Agency (ESA). The original Cassini1 benchmark is a single-objective problem and frequently used as case study for global optimization algorithms due to its highly nonconvex and very sensitive objective function. Here, the problem is extended to four objectives and thus classified as many-objective problem. The MIDACO optimization software represents an evolutionary hybrid algorithm and is used to solve the considered application in regard to two aspects. The first aspect considers the impact of massively parallelized co-evaluation in regard to reaching the global optimal solution and its influence on the solution objective space (particular the Pareto front shape). As a second aspect, the impact of a varying BALANCE parameter, which controls how much importance is given to each individual objective within a multi-objective preference scheme recently introduced as Utopia-Nadir-Balance, on the Pareto front shape is given. In regard to the first aspect, the results show that massive parallelization is an effective remedy to reduce the notoriously high number of sequential function evaluation while still maintaining a sufficient well distributed Pareto front. In regard to the second aspect, the results indicate that an exclusive focus on the first objective is preferable over a BALANCE parameter which distributes the preference over several objectives for this very special kind of application.