SPONSOR: National Science Foundation

BUDGET: $410,670

ABSTRACT: Traffic congestion and the associated delay and economic costs it causes are a source of significant concern. In the United States over the past twenty years, vehicle miles traveled for passenger cars grew 44%, but interstate miles increased less than 8%! In response, transportation departments are moving towards intelligent transportation management through the use of tools such as adaptive ramp meters and traffic signals and expanded traffic information systems. Much of the data available for use in intelligent transportation management is in the form of data streams, such as inductive loop detector data, Automatic Vehicle Location (AVL) systems on buses, and live traffic signal data. This situation presents a clear opportunity for the use of Data Stream Management Systems (DSMS) in Intelligent Transportation Systems (ITS). Several DSMS prototypes and many algorithms for data stream processing have been developed in recent years. However, an examination of the usage of existing DSMS in ITS, clearly demonstrates weaknesses in current DSMS technology that need to be addressed before DSMS can be used in ITS applications. ITS data has characteristics that have received limited study in current DSMS; it is disordered, dirty and bursty and can arrive from widely varied sources (loop detectors, active vehicles, passive transponders). Further, the live data must be compared with archived historical data and other types of information sources. In addition to the current focus on temporal aggregation, ITS data requires spatial and potentially spatio-temporal aggregation. Finally, any DSMS that processes ITS data must scale to thousands or tens of thousands of simultaneous queries. The goals of this research are to extend the Niagara streamprocessing to accommodate queries that arise in intelligent transportation management and information systems (particularly those combining live and archive data), develop improved evaluation techniques that will match transportation applications and data in speed and scale, and then thoroughly test and e valuate the results using the live and archival data sources available in the ITS lab.

The assembled team brings nationally-recognized expertise in data-stream processing and intelligent transportation systems to the project. The research on stream processing builds from the work to date on the Niagara stream-processing system (and by other groups). Further, we have unique resources available to this research. In particular, the ITS lab has direct access to live loop detector data and weather data from the Oregon Department of Transportation (ODOT), soon to be supplemented with incident reports, as well as an archive of this data. This research will develop new stream processing technologies to allow real-time processing of streaming and historical data and develop new techniques for windowed operators. By studying their application of DSMS to a live ITS system, the project will advance the knowledge and understanding of both DSMS systems and ITS research.

The broader impacts of this proposal range from the development of educational materials that will be used for tutorials both for intelligent transportation systems and data management systems to benefits to the ITS community and society at large via improved traffic management technology. The Niagara stream-processing software will be deployed in the ITS lab and will provide infrastructure for further joint projects exploring real-time adaptive traffic management strategies and also for student projects. Applying DSMS technology to transportation data will support more rapid development of experimental and operational intelligent traffic management applications, and make those applications more scalable and easier to modify. The software developed will be made available to the ITS lab and its collaborators. This project will also aid in the training and recruitment of computer scientists and transportation engineers (and the cross-training of both), through involvement of graduate and undergraduate students in the research, use in the classroom of the data sources and software developed and hosting of high-school interns through PSU's Saturday Academy.