Distributed Hybrid Metaheuristics for Optimization

A metaheuristic is an intelligent, iterative process that guides a search and can be applied towards optimization problems, such as the Traveling Salesman Problem. Two well studied techniques for solving optimization problems are Genetic Algorithms and Ant Colony Systems. However, each metaheuristic has different strengths and weaknesses. Genetic Algorithms are able to quickly find near optimal solutions, but may end up in local optima. Ant Colony Systems offer an approach that considers a larger portion of the search space, but require well tuned parameters to compete with the convergence rates of genetic algorithms. Traditional heuristics for solving the Traveling Salesman Problem may perform better than these metaheuristics, but require specific knowledge of the problem domain. Using metaheuristics to solve optimization problems allows for generalized problem solving and may be applicable to different problem domains. This paper introduces GoatSpot, a hybrid approach for solving the traveling salesman problem using distributed genetic algorithms and ant colony systems. GoatSpot is used to analyze the performance of a solution sharing, hybrid metaheuristic.