CAREER: Sensor Fusion for Multi-Modal Traffic Sensing (NSF Award CNS-1149989)

This is a project page supplementing the annual project reports for NSF Award CNS-1149989 specifically. Additional information about work performed by my group, as supported by multiple grants and agencies, may be found under the main lab page.


The objective of this project is to lay the foundation for comprehensive real-time mon- itoring of traffic conditions on urban streets and highways. A scalable, model-based sensor fusion system is proposed to merge data from a large number of heterogeneous and geo- graphically diverse sensors, and to maintain a coherent view of sensed, and inferred, traffic conditions throughout the city.

A city-wide, detailed and accurate view of current traffic conditions will enable a wide range of current and future applications in transportation, including traffic aware route planning, adaptive signalized intersections, road use fees, congestion tolling, law enforcement, and traditional traffic engineering.

Today, traffic sensing on urban freeways is relatively commonplace. Due to their simple topology, accurate estimates on controlled-access highways can be achieved with a relatively small number of sensors. Surface streets, however, present a more challenging problem. Ex- isting approaches, based on cell-tower handoff records or crowd-sourced GPS probe vehicles, lack the accuracy (in the former case), and density (in the latter case), to offer a reliable service away from the freeway corridors.

Through collaborations with the City of Chicago, NAVTEQ, GCM Travel and the Chicago Transit Agency, the PI has access to Chicago-area data from sensing modalities including underground magnetic loops as well as bus, city-vehicle and crowd-sourced GPS traces. To complement these sources, two additional traffic sensing modalities are investigated in this project: video-based vehicle counting, re-identification and classification using in-situ video feeds (using City of Chicago traffic and security cameras), and WiFi-based vehicle re-identification for travel time measurements, using in-situ connectivity and/or equipment.

Project Personell

Publications Resulting from Project

James Biagioni and Jakob Eriksson. Map Inference in the Face of Noise and Disparity. In GIS, pages 79--88. ACM SIGSPATIAL, November 2012. [ .pdf ]
Xuemei Liu, James Biagioni, Jakob Eriksson, Yin Wang, George Forman, and Yanmin Zhu. Mining Large-Scale, Sparse GPS Traces for Map Inference: Comparison of Approaches. In KDD, pages 669--677. ACM, August 2012. [ .pdf ]
Copyright 2016 The Board of Trustees
of the University of
Helping Women Faculty Advance
Funded by NSF