基于再励学习蚁群算法的多约束QoS路由方法  

A Reinforcement Learning Based Ant Algorithm for Multiple Constrained QoS Routing Problem

作　　者：陈岩[1] 杨华江[1] 沈林成[1] 

机构地区：[1]国防科技大学机电工程与自动化学院,长沙410073

出　　处：《计算机科学》2007年第5期25-27,44,共4页Computer Science

基　　金：973基础研究项目(No.5130801)

摘　　要：本文研究了多约束QoS路由问题,给出基于模糊评判的路由模型,实现了多QoS约束的综合优化;同时提出一种再励学习蚁群路由算法对该问题进行求解,算法通过对蚂蚁搜索路径进行评价产生再励信号,并根据再励信号采取了不同的信息素更新策略,提高了算法的寻优能力和收敛速度。仿真实验表明,该算法能快速得到较大程度满足业务QoS要求的路径。This paper discusses the multiple constrained QoS routing problem. Firstly, a mathematical model based on fuzzy judgment is presented, which realizes the optimization of multiple constraint of QoS. Then an Ant algorithm is proposed to solve the problem. An efficient reinforcement learning mechanism, which improves the pheromone according to the reinforcement signal generated from the judgement of the routes, is introduced to the algorithm, so that the algorithm can converge to the approximate global best solution fast. Simulation results demonstrate that the algorithm can effectively and fast generate a route which can mostly satisfy the QoS constraints of operations.

关 键 词：多约束QOS 模糊评判 网络路由 再励学习 蚁群算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]