The path inference filter: model-based low-latency map matching of probe vehicle data

TL;DR

The paper introduces the path inference filter (PIF), a real-time, model-based algorithm for map-matching sparse GPS data, improving accuracy and scalability in diverse urban environments.

Contribution

It presents a novel, efficient framework for real-time map matching that generalizes prior methods and includes automatic training procedures for new data without ground truth.

Findings

Outperforms existing map-matching algorithms on large datasets

Enables real-time processing of sparse GPS data

Deployed at industrial scale in multiple cities

Abstract

We consider the problem of reconstructing vehicle trajectories from sparse sequences of GPS points, for which the sampling interval is between 10 seconds and 2 minutes. We introduce a new class of algorithms, called altogether path inference filter (PIF), that maps GPS data in real time, for a variety of trade-offs and scenarios, and with a high throughput. Numerous prior approaches in map-matching can be shown to be special cases of the path inference filter presented in this article. We present an efficient procedure for automatically training the filter on new data, with or without ground truth observations. The framework is evaluated on a large San Francisco taxi dataset and is shown to improve upon the current state of the art. This filter also provides insights about driving patterns of drivers. The path inference filter has been deployed at an industrial scale inside the Mobile Millennium traffic information system, and is used to map fleets of data in San Francisco, Sacramento, Stockholm and Porto.