THE ADVANCE TARGETED DEPLOYMENT: EVALUATION AND LESSONS LEARNED

The Advanced Driver and Vehicle Advisory Navigation Concept (ADVANCE) advanced traveler information system (ATIS) demonstration project in northeastern Illinois was re-scoped in late 1994 from its originally planned deployment of 3,000 - 5,000 in-vehicle navigation units to a "targeted" deployment in which up to 75 vehicles were equipped with devices enabling them to receive real-time traffic information. These devices included 1) global positioning system (GPS) receivers coupled with radio transmission capability that enabled the vehicles while in the ADVANCE study area to serve as dynamic traffic probes as well as recipients of location data; and 2) navigation units that employed a comprehensive map data base and average (static) link travel times by time of day, stored on CD-ROM, which together computed efficient (least duration) routes between any origin and destination in the northwest portion of the Chicago metropolitan area. Experiments were designed to dispatch these equipped vehicles along links at headways or frequencies comparable to what would have been observed had full deployment actually occurred. Thus, within the limitations of this controlled environment, valuative experiments were conducted to assess the quality of several of the key sub-systems of ADVANCE in the context of structured performance hypotheses. Focused on-road tests began on June 1 and continued through December 14, 1995, followed by a period of data evaluation, documentation of results, and development of conclusions about the findings and usefulness of the project. This paper describes the tests; reports on issues, data collection and analysis problems that arose during their execution (and their resolution); and presents findings and conclusions about test results that are known at time of writing. Among key findings are that over 99% of the 50,000-plus probe reports collected during focused tests of the travel-time prediction subsystem were determined to be reliable and usable by the prediction algorithm; that as few as three probe traversals per 15-minute interval on a given (arterial) link can generate travel time predictions of reasonable accuracy for that link; that the ADVANCE system occasionally but not frequently provides alternative point-to-point routings with the capability to save driving time in an arterial network; and that drivers residing in the study area who were given two weeks to use the system on a daily basis provided very useful insights into what the look, feel, and capabilities of a successfully commercialized route guidance system should be.