| The focus of this thesis is to compare the performance of two optimization techniques: Particle Swarm Optimization (PSO) and Evolution Strategy (ES). Both algorithms were first implemented, tested on a set of benchmark problems, and both were applied to the soccer kick application. The results were visualized, analyzed, and interpreted, which allowed us to determine which of the two algorithms is best suited for these types of problems. Specifically, the dynamics of the PSO algorithm are explored and the effects of various inherent parameters are discussed. We found that particle swarms provided better results (in terms of accuracy and efficiency) in comparison to evolution strategies both on the benchmark functions and the real-world application based on the experiments conducted so far. Particle swarms also provided us with further insights on the soccer kick simulation by finding a subset of values (for the 54 parameters) that yielded good solutions. (Abstract shortened by UMI.). |
