Global Optimisation Heuristics and Test Problems for Preliminary Spacecraft Trajectory Design

Abstract

We describe a number of global optimisation prob- lems representative of spacecraft trajectory design. Each problem is transcribed in the form of a blackbox objective function accepting, as inputs, the decision vector and returning the objective function and the constraint evaluation. The computer code is made available on line as a challenge to the community to develop performing algorithms able to solve each of the problems proposed in an efficient manner. All the problems proposed draw inspiration from real trajectory problems, ranging from Cassini to Rossetta to Messenger to possible future missions. We report the results coming from the application of standard global optimisation algorithms, with unoptimised default settings, to each one of the problems. We consider Differential Evolution, Particle Swarm Optimisation, Genetic Algorithm and Simulated Annealing. These standard implementations seem to fail to solve complex problems in a short time. We conclude the paper introducing what we call “cooperative” approaches between the different algorithms and we show how, already simple cooperation strategies, allow to improve the trajectory optimality.