信号交叉口排队长度预测的神经网络方法

预测信号交叉口的排队长度可以为交通信号控制和管理提供非常重要的信息。基于神经网络，针对定时和感应信号交叉口两种不同情况，成功实现了单、双车道排队长度的预测。同时，感应器与停车线之间的距离对预测精度的影响也进行了初步分析。模拟结果同时表明，神经网络对左转车道排队长度的预测效果不佳，不能为信号控制和管理提供有效的信息。

Prediction of Traffic Queues at Signalized Intersection Using Neutral Networks

Prediction of traffic queues at signalized intersection provides important information for traffic signal control and management. Using neural networks, the study in this paper attempts to predict the traffic queues with a great success for both one and two lanes at both timed and actuated signalized intersections. In addition, the effects of the distance between the detector and stop line on the prediction results are analyzed. However, results of the study indicate that the method can not properly predict the traffic queues at left-turn, and therefore, can not provide valuable information.