摘要
针对当前各种路由算法在广域网环境下由于不能适应各种拓扑环境和负载不均衡时所引起的路由性能不高等问题,提出了一种基于梯度上升算法实现的增强学习的自适应路由算法RLAR。增强学习意味着学习一种策略,即基于环境的反馈信息构造从状态到行为的映射,其本质为通过与环境的交互试验对策略集合进行评估。将增强学习策略运用于网络路由优化中,为路由研究提供了一种全新的思路。对比了多种现有的路由算法,实验结果表明,RLAR能有效提高网络路由性能。
Aimed at the poor performance of the current various routing algorithms,due to the poor adaptability to various changing net-work topologies and loads,an adaptive routing algorithm called RLAR is proposed,and the algorithm is based on reinforcement learning which implemented by gradient ascent algorithm.Reinforcement learning means learning a policy that a mapping of states into actions which based on feedback from the environment.The learning can be viewed as browsing a set of policies while evaluating them by trial through interaction with the environment.Applying the reinforcement learning strategy to the research of routing,as a novel method,the theory is proved.The performance of RLAR and other routing methods is comprehensively compared,lots of simulation results show that RLAR can remarkably enhance the performance of network routing.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第4期1190-1194,共5页
Computer Engineering and Design
基金
国家973重点基础研究发展计划基金项目(2005CB321801)
国家自然科学基金项目(60873215
60621003)
高等学校博士学科点专项科研基金项目(200899980003)