Abstract: Focused on the unsaturation of traffic flow at isolated urban intersection, an isolated intersection self-adaptive control strategy based on K-nearest neighbor short-term traffic flow prediction method was proposed to make the isolated intersection control strategy applicable for dynamically variable traffic flow. Firstly, the traffic flow of the future 5min was predicted according to the historical traffic flow data at the intersection, and the signal cycle was calculated. Then, a self-adaptive control model of the isolated intersection with parameters of maximum cycle, phase sequence and maximum phase time was established, which could adjust automatically the parameters value of phase sequence, the maximum phase time and green light delay time according to the real-time traffic state information. Finally, an isolated intersection in Nansha District of Guangzhou was simulated by software VISSIM 4.3, and the average vehicle delay time was taken as the evaluation index, which was compared with the fixed timing and the piecewise timing control methods. The simulation results show that compared with the fixed timing and the piecewise timing control method, the total delay time of the vehicle obtained by the proposed method is reduced by 35% and 15% respectively, which effectively improves the vehicle traffic efficiency at the intersection.

Key words: traffic and transportation engineering, urban traffic, traffic flow prediction, K-nearest neighbor algorithm, self-adaptive control, isolated intersection

摘要： 针对城市孤立交叉口交通流量未饱和的情形，为了使单交叉口控制方案适应于动态变化的交通流，提出了一种基于K近邻短时交通流预测的单交叉口自适应控制策略。首先依据路口历史交通流量数据预测未来5 min的交通流量，计算信号周期；然后建立以最大周期、相序、最大相位时间为参数的单交叉口自适应控制模型，根据实时的交通状态信息自动地调整相位顺序、最大相位时间和绿灯延长时间；最后使用VISSIM4.3软件，对广州市南沙区某单交叉口进行仿真，以车辆平均延误时间作为评价指标，与固定定时和分段定时控制方法相比较。仿真结果表明：本文方法的车辆总平均延误时间与固定定时和分段定时控制方法相比分别减少35%和15%，有效提高了交叉路口的车辆通行效率。

关键词: 交通运输工程, 城市交通, 交通流预测, K近邻算法, 自适应控制, 单交叉口