基于SA-FQL算法的区域交通控制方法  

Urban traffic control based on SA-FQL

作　　者：邓军[1] 刘智勇[1] 

机构地区：[1]五邑大学信息学院,广东江门529020

出　　处：《计算机工程与应用》2010年第27期231-233,237,共4页Computer Engineering and Applications

基　　金：广东省自然科学基金No.8152902001000014;广东省高等学校自然科学重点研究项目(No.05Z025)~~

摘　　要：将模拟退火算法的Metropolis准则用于平衡模糊Q学习中探索和扩张之间的关系,提出基于Metropolis准则的模糊Q学习算法Simulated Annealing Fuzzy Q-learning(SA-FQL)。利用SA-FQL算法优化区域的公共周期,在给定周期的基础上再用SA-FQL算法优化区域中各干线相邻两路口的相位差,最后根据交通流量确定各路口的绿信比。TSIS仿真结果表明,相比基于Q学习和模糊Q学习的控制方法,该方法能显著提高学习速度和交通效率。For area traffic control,the Simulated Annealing Fuzzy Q-Leaming（SA-FQL）,a modified version of the popular Fuzzy Q-Learning, is used.Firstly, this paper optimizes the common cycle of the traffic network using SA-FQL, then optimizes offset for every trunk road using the same method based on the common cycle, lastly adjusts split for every intersection according its traffic volume.TSIS simulation results show that, compared with Fuzzy Q-Learning method and Q-Learning method,the method proposed in this paper can significantly accelerate learning and improve traffic efficiency.

关 键 词：区域交通控制 模拟退火模糊Q学习算法 模糊Q学习 Q学习 METROPOLIS准则 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]