摘要
为了检测Q学习算法在信号控制方案中的效果,在Webster配时法的基础上,建立了适应交通信号控制及以车均延误最小为目标的奖惩函数,并详细说明了Q学习独立交叉口信号控制的原理和应用过程.通过流量波动大和小两个算例,验证了Q学习控制优于定时控制.
The goal of the paper is to detect the effect of Q-learning method on traffic signal control. Based on Webster's timing algorithm, a reward-penalty function adaptable to traffic sig- nal control for the minimization of average delay is estabilished. It also illustrates the control theory and application process of Q-learning signal control in single intersection. From the two examples in large and small flow fluctuation, the author verified that Q-learning control is better than fixed-time control.
出处
《交通科学与工程》
2009年第3期90-94,共5页
Journal of Transport Science and Engineering
基金
湖南省教育厅科研资助项目(09A003)
长沙理工大学公路工程省部共建教育部重点实验室开放基金资助项目(kfj080102)