Abstract: Traffic light with countdown timer is widely used today, this paper introduces an efficient but simple control algorithm. This algorithm analyzes the statistics of vehicle passing rates of each lane near traffic lights and classifies them by online clustering, and evaluates optimal solutions of each category by genetic algorithm. It aims at sufficient category refinement and solution optimization. Theoretically, it can allot green light time for each lane proportional to their vehicle densities. Simulation verifies its effectiveness in reducing average waiting time of vehicles, compared with other algorithms supporting counting-down timers.

Key words: traffic lights with countdown timer, genetic algorithm, clustering, vehicle density

摘要： 针对目前被广泛使用的具有倒计时功能的交通灯，设计一种简洁高效的控制算法。它实时对路口各车道车流量数据进行统计分析并聚类，每个交通灯周期都根据当前路况执行相应路况类别下的遗传算法过程，以寻求该类别的最优解。算法目标是实现分类的足够细化和各类别解的足够优化。测试表明，该算法能根据车道的车辆密度合理分配各车道的绿灯时间，使车辆的平均等待时间少于其他支持倒计时功能的算法。

关键词: 倒计时交通灯, 遗传算法, 聚类, 车辆密度

CLC Number: 

• U491.51