Traffic congestion has a significant impact around the world. Building reliable and cost effective traffic monitoring systems is a prerequisite to addressing this phenomenon. Historically, traffic estimation has been limited to highways, and has relied on a static, dedicated sensing infrastructure such as loop detectors or cameras. In the case of city roads, this estimation problem is rather involved. This situation can be partly attributed to the lack of effective sensing in an urban setting. In this context, the most promising source of data is the GPS receiver in personal smartphones and commercial fleet vehicles. In this article, we present some algorithms that leverage this trend to produce some streaming data compatible with current state-of-the-art traffic estimation algorithms. These algorithms, which we will refer altogether as the path inference algorithm, have been implemented and deployed inside the Mobile Millennium system at Berkeley.

本篇论文介绍了一种叫做路径推断滤波器(PIF)的新型算法，用于在稀疏GPS点的数据下从中推断出车辆的轨迹，并可通过真实数据进行高效训练，通过在大量数据集上的评估得出比当前最先进的方法更为精确，还通过此过滤器获得了有关驾驶员驾驶模式的见解。此算法已经在Mobile Millennium交通信息系统中工业规模部署，在旧金山、萨克拉门托、斯德哥尔摩和波尔图等地映射了大量的数据集。

基于模型的低延迟路线推断过滤器：探针车数据匹配地图