TO1 - TO2: Scope of work and current status

Background: Caltrans owns, operates, and maintains fixed sensors throughout the state to gather necessary data for (1) planning, (2) traveler information systems, and (3) traffic operations. We believe the Caltrans data pipeline of the future will also include probe data derived from GPS devices embedded in smartphones, personal navigation devices, taxis, fleets and other sources.

Goals: The TO1-TO2 contract specified a number of institutional, practical, and analytical goals. The first step was to understand how to purchase data, to define data quality, to develop bid documents, and to negotiate a contract. The second step was to gain experience in handling the data and in building real-time tools and processes for data storage, quality analysis, validation, and fusion. The third step was to advance the science: to integrate both probe and fixed sensor data through a flow model, to determine the value of probe data for the purpose of travel time estimation, and to measure the relationship between probe data coverage and travel time estimation accuracy.

Accomplishments: Thus far, success has been demonstrated in all three goal areas. Contracts were awarded for a 90-days data procurement from each of two vendors. Three sites chosen for evaluation were (1) I-880 near Hayward and Fremont, (2) an urban area along I-15 in Ontario, and (3) a rural area without loop detectors along I-15 in Victorville. Tools and systems were successfully developed for data storage, quality analysis, validation, and fusion. In addition, and for each site, 10 Bluetooth readers were installed along the highway in distances of roughly 1-2 miles. The Bluetooth data were collected, filtered, and used as an independent verification of vehicular travel times. Automated flow-model calibration algorithms were implemented. Initial data assessments revealed good agreement among each of the data sources.

Ongoing Work: Work continues to determine the level at which to trust individual data sources for the best fusion result, and to predict bounds on travel time estimation accuracy as a function of available data.