SBIR Phase I: Mapping Driver Behavior: Toward an Active Safety Knowledge Base

The innovation put forth in this proposal consists of a database that captures and describes expected behavior of vehicles at any point of a road network. The range of driving behaviors on a particular road, e.g. speed and lateral position, can serve as a reference for other drivers and systems to increase driving safety and efficiency. Deviations from the normal range may indicate a risky driver, training opportunities, or the need for immediate intervention from on-board driver assistance systems. The project brings together two innovative features. The first one is the use of vehicle data as a primary source to build a digital map. The second feature lies in representing behaviors rather than the physical road structure that current digital maps focus on. This has significant advantages for the intended applications. To enable this feature, this project will demonstrate a novel core map architecture that allows the accurate capture of maneuvers and behaviors. This is nearly impossible using currently available map formats. In this project a simple driving behavior database will be built and will validate that the proposed data representation can be effectively constructed and scaled to cover large geographic areas. The broader/commercial impact of the proposed innovation will be significant gains in driving safety and efficiency. Capturing behaviors allows their analysis to assess risk and identify optimal maneuvers. This enables feedback to drivers to either nudge or actively pull them toward safer and more efficient patterns. This database will address needs and applications in several transportation markets: 1) driver assistance systems triggering personalized warnings based on nominal maneuver parameters and driver?s habitual style; 2) risk assessment for fleets and insurance companies- identifying outlier drivers, providing training/feedback and setting insurance rates; 3) providing inputs to autonomous vehicle systems to set trajectories that are comfortable for humans and are understood by other drivers; 4) roadway safety for infrastructure managers- observing perturbations in vehicle movements to diagnose problems and engineer remedial upgrades; and 5) identifying driving practices that reduce fuel usage and improve the range of electric vehicles. These are large target markets with considerable societal value and revenue potential. The proposed database can be built automatically from vehicle data alone, significantly reducing costs compared to existing map databases. Novel and valuable roadway and driving attributes can also be captured.