基于简化路网模型的卡尔曼滤波多步行程时间预测方法

针对目前多步行程时间预测方法研究较少,存在未来一段时间内的观测值不能及时得到的问题,提出基于简化路网模型的卡尔曼滤波多步行程时间预测模型和算法.综合运用上游路段、当前路段的实时和历史行程时间数据,从历史数据中寻找与当前日期交通状况最接近的历史日期,解决卡尔曼滤波未来一段时间内没有观测值而无法进行多步预测的问题.实验表明,算法能够合理地预测未来几个时段的路段行程时间,有效地避免了时滞性.同时,多步行程时间预测算法的精度高(尤其是4步以内,均等系数达到0.9以上), 是一种可行的预测方法.

Multi-step Kalman filtering travel time estimation method based on simplified road network model

Multi-step travel time prediction is more reasonable in reality than one-step travel time prediction method, but it was done research in by few scholars. Meanwhile, the real-time travel time in future several time sections cannot be detected. In this paper, multi-step travel time estimation model and algorithm based on Kalman filtering (KF) were presented. This model took real-time and history travel time of the upper road and road itself into account. The algorithm found the nearest time in history data away from today. They can solve the problem that travel time in future are not be detected, and travel time in future several time sections can be predicted reasonably. Meanwhile, the time lag was solved effectively. Results of experiments show that the accuracy of several-step prediction algorithm is great, and it is meaningful in practical application.