Project: Automated identification of traffic patterns

Principal investigator. Dr. Brian L. Smith, University of Virginia, 804-243-8585, briansmith@virginia.edu.

External contact. Cathy McGhee, VDOT, Virginia Transportation Research Council, Charlottesville, VA. 434.293.1973

Project objective. To develop and evaluate procedures for automated identification of patterns from large traffic datasets, and to investigate an alternative paradigm for archiving ITS data.

Project abstract. As part of its Smart Travel program, VDOT collects vast amounts of traffic data through sensors, for use in real-time monitoring and control of traffic, as well as for performance measurement. Archiving this data would be greatly simplified if patterns could be identified. Then a pattern could be noted, and only data which deviated from the pattern would need to be recorded. Identification of such patterns would also be useful for traffic engineering studies, traffic safety studies, transportation planning, and simulation studies. The first requirement in such an effort is to identify basic patterns from a given data archive. This is a challenging exercise, because of the multivariate stochastic nature of traffic data, the lack of a unique basic pattern that can be used for benchmarking, often poor data quality, and the fact that patterns vary with time and location.

This research will extract traffic pattern feature datasets, and compare some competing analytic methods for finding patterns, such as clustering algorithms, wavelet analyses, and spectral time series analyses. Traffic data for I-66 and I-64 archived at the Smart Travel Laboratory at the University of Virginia will be used for evaluating the developed methodology. Data models will be developed for archiving data using just the 'typical patterns' and 'disruptions'.

Tasks. 1) Identify traffic data features dataset. 2) Develop procedures for automated identification of traffic patterns. 3) Evaluate and test procedures on I-64 and I-66 sensors. 4) Develop data models for archiving data with 'typical patterns' and 'disruptions'.

Milestones. Completion by December 31, 2005. (Assumes no-cost grant extension.)

Student involvement. 1 GRA

Budget.

FY04

Faculty 15,000
GRA 5,000

FY05

Faculty 110,000
GRA 10,000

Project total 140,000

Relation to other research. This project will support the ITS Center project "VDOT surveillance needs."

Technology Transfer. Papers in professional journals and presentations at conferences

Potential Benefits. Improved Archived Data User Service. Improved 'typical traffic' information available for traffic/safety analyses.

TRB Keywords: typical traffic, pattern recognition, wavelets, spectral analyses, data archive, ITS