A behavioral-based meta-heuristic for robust global trajectory optimization

This paper presents a behavioral-based meta- heuristic for black-box problems of global trajectory optimization. This approach is shown to perform an efficient exploration of the solution space without sacrificing local convergence. The proposed meta-heuristic models the search for a solution as an action-selection process: a number of agents, forming a population, are endowed with a number of individualistic and social behaviors. The combination of these behaviors drives the entire population toward a number of local optima and eventually to the global one. In order to improve the collection of local optima in different regions of the search space the behavioral-search has been hybridized with a domain decomposition technique. This approach was applied to two typical problems in trajectory design, demonstrating a remarkable robustness compared to the most common methods, both stochastic and deterministic, for global optimization.