Authors: Yiwen Zhong; Lijin Wang; Changying Wang; Hui Zhang
Addresses: College of Computer and Information Science, Fujian Agriculture and Forestry University, Fuzhou, 350002, China. ' College of Computer and Information Science, Fujian Agriculture and Forestry University, Fuzhou, 350002, China. ' College of Computer and Information Science, Fujian Agriculture and Forestry University, Fuzhou, 350002, China. ' Pervasive Technology Institute, Indiana University Purdue University Indianapolis, Indianapolis, IN 46202, USA
Abstract: Simulated annealing (SA) algorithm is extremely slow in convergence, and the implementation and efficiency of parallel SA algorithms are typically problem-dependent. To overcome such intrinsic limitations, this paper presents a multi-agent simulated annealing (MSA) algorithm to address continuous function optimisation problems. In MSA, a population of agents run SA algorithm collaboratively, exploiting the mutation operator formulas of differential evolution (DE) algorithm for candidate solution generation. Our MSA algorithm can achieve significantly better intensification ability by taking advantage of the learning ability from DE algorithm; meanwhile the probability acceptation rule of SA algorithm can keep MSA from premature stagnation. The MSA algorithm is population based, so it can be paralleled problem-independently and easily. Simulation experiments were carried on four typical benchmark functions, and the results show that MSA algorithm has good performance in terms of convergence speed and solution accuracy.
Keywords: multi-agent simulated annealing; MSA; differential evolution; continuous function optimisation; intensification; bio-inspired computation; multi-agent systems; MAS; agent-based systems; simulation.
International Journal of Bio-Inspired Computation, 2012 Vol.4 No.4, pp.217 - 228
Published online: 18 Jul 2012 *Full-text access for editors Access for subscribers Purchase this article Comment on this article