基于神经网络与群智能的多代理人决策模型及其应用.pdfVIP

第 18卷 第 1期 运 筹 与 管 理 V o.l 18, No. 1 2009年 2月 OPERAT IONS RESEARCH AND M ANAGEM ENT SCIENCE F eb. 2009 1 2 1 黎 明 , 刘伟兵 , 王先甲 1. , 430072; 2. , 430072) : , , , , P are to : ; ; ; : C934 : A : 1007-3221 2009) 01- 0007-05 M ult-i agen t D ecision M ode l and A pp licat ion Based on N eura l N etw ork and Sw arm Intelligence 1 2 1 L IM ing , L IU W e-i b ing , WANG X ian-jia ( 1. Institute of System s E ng ineering, Wuhan University, W uhan 430072, China; 2. School of P olitical Science and PublicM anagem ent, W uhan University, W uhan 430072, China Ab stract: T h is paper proposes a mu lt-i agen t dec ision m odel based on neura l newt orks and sw arm in telligence techno logy. In th is paper, a neura l network is u sed for behav ior dec ision contro ller. The inpu ts of the neural net- w ork are dec ided by the last actions o f other agents. T hen the outpu ts determ ine the nex t action that the agent w ill choose. T he w e igh t values are updated by sw arm inte lligence opt mi ization algorithm, and they mi ply the be- hav ior evolution o f agents. The validity of the dec ision m odel is verified through smi u lation exper mi ent, and the resu lts show that th is decision m ode l has the ab ility of adaptive learning and can prevent the co llision bewt een a- gents to obta in the P areto opt mi al. Key wo rds: m