Emergent search on double circle TSPs using subgour exchange crossover

Genetic algorithms (GAs) have such potentials for realizing emergent searches that local search techniques, such as simulated annealings, and parallel local search techniques, like other evolutionary computation such as evolution strategies and evolutionary programmings, do not have. Crossover operators bring these potentials because they can emerge their neighborhood structures as populations evolve. The paper presents a realization of emergent searches by GAs. First, we show difficulties for local search techniques to solve double circle TSPs, and discuss how emergent searches can overcome such difficulties. Second, we propose guidelines to achieve emergent searches by GAs; the characteristics preserving encodings/crossovers design and the diversity maintaining generation alternation models design. According to these guidelines, we actually realize GAs to solve double circle TSPs