Genetic Particle Swarm Optimization Based on Multiagent Model for Combinatorial Optimization Problem

Particle swarm optimization can be viewed as a distributed agent model, but many agent computing characteristics are still uncovered. This paper combines multiagent system and genetic particle swarm optimization (GPSO) and proposes a multiagent-based GPSO approach (MAGPSO), for combinatorial optimization problems. In MAGPSO, an agent represents a particle to GPSO and a candidate solution to the optimization problem. All agents live in a lattice-like environment, with each agent fixed on a lattice point. In order to obtain optimal solution quickly, they compete and cooperate with their neighbors, and they can also use knowledge. To demonstrate its performance, experiments are carried out on a combinatorial optimization problem, bipartite subgraph problem. The results show that the proposed algorithm has superior performance to other discrete particle swarm algorithms by using the agent- agent interactions and evolution mechanism of GPSO in a lattice-like environment.