一种新的免疫协同多Agent模型及其仿真分析  

Novel Immune Cooperative Multi-agent Model and Its Simulation Analyses

作　　者：钱海[1] 马建辉[1] 王煦法[1] 

机构地区：[1]中国科学技术大学计算机科学技术系,安徽省计算与通讯软件重点实验室,合肥230027

出　　处：《系统仿真学报》2008年第13期3436-3439,3444,共5页Journal of System Simulation

摘　　要：提出了一个具有高效问题求解能力的多Agent系统模型。基于生物免疫系统的自适应识别机制,将基因库进化与亲和度成熟、元动力学等要素相结合,设计具有自适应性和强大搜索能力的单Agent结构,多个Agent基于形态空间模型协同演化并涌现出智能求解能力。模型采用基于群体的多点随机搜索以及多Agent完全并行的执行方式,是一种新型协同演化模型。仿真实验结果证明了模型的有效性。A novel multi-agent system model that could be used for efficiently solving practical problems was proposed. Based on the adaptive immune recognition mechanism of the natural immune system, the architecture of single agent was designed by combining gene library evolution with key factors such as affinity maturation and meta-dynamics. According to the shape space model, all agents in the model co-evolve and emerge intelligent problem-solving ability. This is a novel co-evolution model which uses population based multi-point random search, and all agents execute in parallel. Results of simulation experiments prove the availability of the new model.

关 键 词：多AGENT系统 人工免疫系统 协同演化 克隆选择 K-旅行商问题 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]