| In response to growing concern over traffic congestion, traffic management systems have been built in large urban areas in an effort to improve the efficiency and safety of the transportation network. This research effort has developed an automated condition monitoring method that utilizes archived traffic data to provide a basis for the assessment of current traffic conditions, and if applicable, a determination of the degree to which current conditions are abnormal. An improvement over commonly used traffic condition monitoring methods is realized in that system state is characterized across a range of conditions rather than as in an incident or incident-free condition. An additional improvement is that rather than analyzing values of mean speed, volume, and occupancy (traffic variables typically measured in a traffic management system) independently, the interrelationships among traffic parameters are exploited using multivariate statistical quality control (MSQC). This statistical approach provides the tools and basis for extensions needed to intelligently assess current traffic conditions using historical data. Prototype applications for use in traffic management systems and for data mining purposes have been developed. These applications employ a newly developed procedure for screening both current and archived data from traffic detectors in order to reduce the potential of using erroneous data in the MSQC-based traffic condition monitoring method. Several strategies for sampling the historical database, based on temporal relationships with current data, have been developed and evaluated. Enhancements to basic MSQC include calculation of the exact degree of abnormality corresponding to T2 (the key measure of normality in MSQC), classification of T2 into degrees of abnormality, and decomposition of T 2 into components for each traffic parameter. The method is evaluated according to measures of traffic condition assessment, method operating characteristics, and detection of accidents. Implementation of the method will allow traffic managers to focus their efforts on abnormally operating locations and then determine an appropriate course of action to attempt to return the system state to normal. Potential benefits include more efficient use of human and computer resources in traffic management centers, improved return on investment in traffic detection infrastructure, and ultimately reductions in traffic congestion. |
