基于门控循环单元模型的在线路网匹配算法

Online map matching algorithm based on the gated recurrent unit model

路网匹配是道路网轨迹数据分析领域的一项关键技术,一个快速且准确的路网匹配算法能够为上层应用提供良好的技术支持.随着轨迹数据的爆炸式增长,现有的在线路网匹配算法存在延时的现象,尤其是在低频轨迹数据的环境下,无法快速地对轨迹数据进行路网匹配.神经网络和深度学习的发展为解决这些问题提供了新的方法.提出了一种利用门控循环单元(Gated Recurrent Unit, GRU)模型快速定位轨迹采样点的候选路段、从而加速在线路网匹配计算的方法,并将此方法和最新的在线路网匹配算法进行了实验比较.结果表明,基于GRU模型的在线路网匹配算法能够有效地加快匹配过程,提高匹配效率.

Map matching is a key technology in the field of road network trajectory data analysis. A fast and accurate map matching algorithm can provide good technical support for upper-layer applications.With the explosive growth of trajectory data, existing online map matching algorithms experience a delay phenomenon;in particular, in the context of low-frequency trajectory data, it is impossible to quickly perform map matching on trajectory data. The development of neural networks and deep learning provide new methods for solving these problems. This paper uses the gated recurrent unit(GRU) model to quickly locate candidate segments of trajectory sampling points, thus accelerating the calculation process for online map matching. The proposed method is experimentally compared to the latest online map matching algorithm;the results show that the GRU-based online map matching algorithm can effectively speed-up the matching process and improve matching efficiency.