SRVC: Sensornets for Remote Vehicle Classification

The motivation of this research is the need for short-term data collection about freight traffic movement in urban areas. We seek to develop automated vehicle classification systems based on networks of small, battery-powered and wireless, intelligent sensors that can be easily deployed with brief setup time (tens of minutes), with accurate vehicle information (as good as or better than human observers), and communicate this information to a central monitoring site. Current approaches are not rapidly deployable, accurate enough, and lack the ability to relay data in real-time to central site. This report summarizes the results of a two-year research effort on sensornets for vehicle traffic classification. At the study outside, our main goals were to understanding the communications requirements for traffic monitoring systems (both short-range wireless inside a traffic sensornet, and wide-area to a central Traffic Management System), developing self-configuring traffic monitoring systems, and integrating prior work with these new results. The key outcomes of our research are to understand communications in a classification system, to carefully evaluate the effects of signature matching, and to further develop selfconfiguration. Finally, we tested our new approaches through a second major data collection exercise at USC.