Hybrid Multiagent Swarm Optimization: Algorithms, evaluation, and application

Abstract

In this paper, a novel, hybrid-combined swarm optimization algorithm which consists of both particle swarm optimization (PSO) and multiagent coordination optimization (MCO) is proposed. Such a new algorithm is named after Hybrid Multiagent Swarm Optimization (HMSO). The standard PSO algorithm, in which the position of each particle approaches the global optimal solution with certain velocity, is effectively solving non-consensus-type optimization problems, that is, the components of the optimal solution vector are not identical. While the MCO algorithm, in which the agents can share their information to achieve consensus with a fast convergence speed, is effectively solving consensus-type optimization problems. The main difference between the MCO and PSO algorithms lies in the local optimization solutions. Hence, the HMSO algorithm combines both advantages from the PSO and MCO algorithms to solve either consensus- or non-consensus-type optimization problems more efficiently. In particular, the HMSO algorithm achieves the optimal value which is close to the better one between two optimal solutions obtained by the PSO and MCO algorithms for solving consensus-type and non-consensus-type optimization problems, respectively. This point has been shown by the comparison of these three algorithms with eleven standard test functions. Finally, an application example is solved numerically by using these three algorithms, in which HMSO shows the best result among the three algorithms.