Fitness Landscape Analysis for Spacecraft Trajectory Optimizations

This paper presents fitness landscape analysis (FLA) for investigating fitness landscape characteristics and problem difficulty for spacecraft trajectory optimization problems. The FLA is performed to understand/identify structural characteristics of the optimization problem. Five FLA metrics are used to quantify the problem features. They can measure not only the fitness landscape shape such as ruggedness and neutrality but also identify problem properties such as modality and funnel structure. We apply those metrics to spacecraft trajectory optimization problems and investigate the fitness landscape features in detail. In addition, the correlation between the algorithm performance and FLA metrics is analyzed, and appropriate FLA metrics for predicting problem difficulty are proposed.