基于ST-GCN短时路况预测算法的预警系统

Road Condition Early Warning System Based on Short Term Prediction with ST-GCN

为提升高速公路车速预测的准确性,针对现有车速预测模型存在的时间相关性和空间相关性部分缺失的问题,提出一种基于时空图卷积网络(Spatio-Temporal Graph Convolution Network,ST-GCN)短时路况预测算法的预警系统。该算法综合考虑时间相关性和空间相关性的影响,根据实时和历史的交通数据,通过建立ST-GCN模型分析预测未来某段时间的交通流速度和路况。将预测结果推送给布设在高速公路上的多彩智能情报板,通过多彩智能情报板上显示的信息诱导司乘用户的行为,从而降低事故发生率,提升高速公路通行效率。

Aiming at the problem that time dependence and spatial correlation are not involved in the existing speed prediction models,this paper develops an early warning system based on a short-term road condition prediction algorithm with ST-GCN(Spatio-Temporal Graph Convolution Network)model.This algorithm comprehensively considers the impact of time dependence and spatial correlation.With real-time and historical traffic data,ST-GCN model analyzes and predicts the traffic speed and road condition in a certain period of time in the future.The predicted road condition is pushed to the color display boards installed with the expressway to guide drivers and passengers,thus,reducing the accident rate and improving the traffic efficiency of the expressway.