+CARPS: Configuration of Metaheuristics based on JADE Agents

A novel agent-based approach to the configuration of algorithms including, but not limited to, metaheuristics is proposed in this work. Metaheuristics are examples of algorithms where parameters need to be set up as good as possible. Yet the selection of the most adequate parameter values is a recognized arduous work in the metaheuristics community. This paper proposes a new approach that makes use of agent technology issues to contribute to the configuration of algorithms: the +CARPS architecture. It is a multiagent system that was developed, implemented and tested based on collaborative, FIPA-compliant JADE agents. The agents carry out an optimization process: the process of configuring algorithms through the fine-tuning of their parameters. +CARPS provides the needed infrastructure to support a distributed and concurrent agent-based configuration of metaheuristics. It is also a framework in which monitoring of control factors of metaheuristics becomes an easy task and, at the same time, it is a powerful tool useful for conducting experiments when executing metaheuristic algorithms. The basic features of the architecture are introduced in this paper. Some experimental results are also included.