Hybridisation of metaheuristics is an important subject that has been explored by several researchers. Multiagent systems have been important tools to accomplish the task of hybridising metaheuristics. In those multiagent approaches, however, each metaheuristic is performed separately, and the potential of the hybridisation is not fully explored. In order to bridge this gap, this chapter presents MAHM (multiagent architecture for hybridisation of metaheuristics), a multiagent approach for metaheuristics hybridisation inspired on particle swarm optimisation (PSO). Introduced as a metaheuristic, PSO can be viewed as a multiagent system once particles can be thought as agents that interact and work together to achieve specific goals. In this context, particles are identified to autonomous virtual entities with social ability, reactivity, and pro-activeness. Each agent in MAHM may use different sets of predefined metaheuristics in different moments of the search to look for high-quality solutions of an optimisation problem. Thus, several elements of different metaheuristics can be found running in the swarm every MAHM iteration. In order to show the potentiality of the proposed architecture, computational experiments were carried out with the travelling salesman problem and the quadratic assignment problem, two important test grounds for algorithmic ideas.

Chapter Contents:

• Abstract
• 9.1 Introduction
• 9.2 Related works
• 9.3 The proposed approach
• 9.3.1 Particle swarm optimisation
• 9.3.2 Multiagent architecture for hybridisation of metaheuristics
• 9.4 The travelling salesman problem
• 9.5 Quadratic assignment problem
• 9.6 Computational experiments
• 9.6.1 Experiments with the TSP
• 9.6.2 Experiments with QAP
• 9.7 Conclusions
• References

Inspec keywords: particle swarm optimisation; multi-agent systems; travelling salesman problems

Other keywords: multiagent approach; metaheuristic PSO; multiagent system; PSO-based multiagent architecture; MAHM; predefined metaheuristics; particle swarm optimisation; metaheuristics hybridisation

Subjects: Expert systems and other AI software and techniques; Optimisation techniques