基于行程时间多步预测的实时路径导航算法

针对现有车辆导航算法仅考虑单一数据,使所得路径实际行程时间比预期更长的问题,首先建立了基于卡尔曼滤波理论的行程时间多步预测模型；其次,提出了综合利用实时数据、行程时间多步预测数据及历史数据的实时路径导航算法,并改进了其实现的核心算法Dijkstra_pred.实验结果表明,基于三类数据的实时路径导航算法所得路径的实际行程时间从整体上优于仅采用实时数据的导航算法,且路径变化较少.

Real-time navigation algorithm based on multi-step travel time prediction

In view of the existing algorithms only considering one type of data and longer real travel time than prediction travel time, firstly, this paper proposed multi-step travel time prediction model based on Kalman. Secondly, it presented the route navigation algorithm which based on real-time data, multi-step prediction data and historical data. In the meantime, it designed the core algorithm Dijkstra_pred. Experiments show that the travel time calculated by the proposed route guidance algorithm is superior to the one by the algorithm based on real-time data, and route calculated based on the algorithm proposed in this paper changes less than the one considering only real-time data.