Improving Genetic Algorithm(GA) Needed for Automated Negotiations among Multi-Agents
Hong Zhang,Huacan He
DOI: https://doi.org/10.3969/j.issn.1000-2758.2008.05.014
2008-01-01
Xibei Gongye Daxue Xuebao/Journal of Northwestern Polytechnical University
Abstract:To deal with such problems as long negotiation time, high complexity and occasionally deadlocked negotiation in automated negotiations among multi-agents, we put forward an improved GA in the full paper. We explain our improved GA in some detail in the first section of the full paper. In this abstract, we just add some pertinent remarks to naming the three subsections: automated negotiations among multi-agents and the Pareto optimal solution (subsection 1.1), our improved GA(subsection 1.2) and the performance evaluation of our improved GA(subsection 1.3). In subsection 1.2, we do four things: (1) we describe the evaluation function of our improved GA, using the Pareto optimal solutions; (2) to avoid computing complexity and premature convergence, we select that the population scale between 1.5 to 2 times the coding length of chromosomes' strings; (3) to improve search efficiency, we divide the initial scales of population into two groups, one group for the interests of buyers and the other group for those of sellers; (4) to improve the stability and convergence of the GA, we perform the operation of crossover in population with higher utilities and that of mutation in population with lower utilities. Finally, to compare our improved GA with traditional GAs, we did experiments on many-to-many multi-criteria negotiations in electronic markets. The Pareto optimal solutions from the experiments, discussed in subsections 2.2 and 2.3, show preliminarily that: (1) our improved GA is more effective and efficient than the traditional GA in Ref.1 (authored by Oliveira et al) in that our GA obtains the Pareto optimal solutions every time; (2) the average non-dimensional generation distance is 0.35, much shorter than that of the traditional GA in Ref.2 (authored by Niu).