摘要
针对Q-学习算法收敛慢、易陷入局部最优的缺陷,提出了一种基于灾变模糊Q-学习(CAS-FQL)算法的区域交通协调控制方法,即将灾变策略引入到模糊Q-学习算法的学习过程中,以提高和改进Q-学习的寻优能力和学习效率.具体是,利用CAS-FQL算法分别优化路网中各交叉口的周期和相位差,绿信比则采用常规方法优化.TSIS软件交通仿真的结果表明,相比基于Q-学习的控制方法,CAS-FQL算法能显著加快算法的收敛速度、提高交通效率.
In order to solve the problems of Q-learning algorithm＇s slow convergence and easily running into the local optimum, this paper puts forward a kind of regional transportation coordination controlling method which is based on Catastrophe-Fuzzy Q-Learning （CAS-FQL） algorithm. The catastrophe strategy is combined into the learning process of the fuzzy Q-learning algorithm to enhance and improve its optimization ability and learning efficiency. Concretely, CAS-FQL algorithm is applied to optimize the cycles and offsets of each intersection in the traffic network, and the split is optimized by conventional method. The results from ISIS traffic simulation platform shows that the catastrophe strategy can accelerate the algorithm convergence speed significantly and improving the traffic efficiency.
出处
《五邑大学学报（自然科学版）》
CAS
2012年第3期67-73,共7页
Journal of Wuyi University(Natural Science Edition)
基金
广东省自然科学基金资助项目(8152902001000014)
广东省高等学校自然科学重点研究项目(05Z025)
