The Path Inference Filter: Model-Based Low-Latency Map Matching of Probe Vehicle Data

Abstract

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.

Road Traffic Anomaly Detection via Collaborative Path Inference from GPS Snippets

“…We use static CRF (SCRF) model [8] and HMM [18] model for path inference as baseline algorithms, both adopting the static features such as length of road segment. For our tensor filter, we set ${\lambda }_1$ and ${\lambda }_2$ to 0.01 for Equation (9).…”

“…Under the assumption of Markov independence relations, Hidden Markov Model (HMM) [18] and Conditional Random Fields (CRF) [8] have been explored. Both HMM and CRF need to utilize various features for designing the transition probability of states, which encourage the weight learning algorithm using inverse reinforcement learning [19].…”

“…Previous studies [8,18] usually use static feature function ${\varphi }_s$ with static features such as the length of road segments, the turns of a path, the speed limit of roads, etc. There are two problems in using static features.…”

“…Reconstructing fine-grained trajectories from noisy and sparse GPS snippets, is often referred to as map matching [7] or path inference [8]. However, previous studies often overlook road dynamic context information which is particularly important for path inference.…”

Development of an IGVM Integrated Navigation System for Vehicular Lane-Level Guidance Services

“…Generally, a global positioning system (GPS) receiver is used in the navigation system, and the position data obtained from a GPS receiver is displayed on a graphic user interface (GUI)-based digital map with map matching (Lou et al 2009, Pashaian et al 2012, Hunter et al 2014. In this method, two problems exist: frequent discontinuities of GPS signals in the urban environments, and simple map matching on the center of a road.…”