Vehicle Automation and Transportability of Crash Modification Factors
- Gary Davis, Professor, Civil, Environmental and Geo-Engineering
Safety-based roadway design requires predicting a design's safety consequences, and the primary tools for making these predictions, compiled in the Highway Safety Manual, are based on regression-type statistical models and before/after studies fit to observational data. Like most empirical models these tools can give reasonable representations of the conditions in place when they were constructed, but extrapolating them to new conditions is problematic. In particular, these tools are functions of the driver and vehicle mix prevalent during the last 20 years or so, which are likely to change, possibly drastically, as vehicle automation becomes widespread. A major challenge facing safety researchers will then be to adapt, if possible, the extensive research that went into the Highway Safety Manual to these new and different conditions. Although a considerable and ongoing effort is being devoted to predicting the direct effects of vehicle automation (e.g. Rosen et al. 2010) to my knowledge no attention is being given to how vehicle automation will interact with existing road safety principles. This can be seen as a problem of determining the transportability (i.e. external validity) of estimated crash modification factors to these new conditions. Hauer et al (2012) and Persaud et al (2015) have initiated a discussion of this issue but to date the focus has been on geographical variability under the current driver/vehicle mix. Within the field of artificial intelligence however, Bareinboim and Pearl (2012, 2014) have developed analytic tools which could profitably be applied to this problem. This project will explore the usefulness of Bareinboim and Pearl's tools by looking at how vehicle automation could impact the crash reduction effects of two roadway-based countermeasures (1) installation of pedestrian hybrid beacons (PHB) at uncontrolled crosswalks and (2) offsetting opposing left-turn lanes at signalized intersections.
- Start date: 02/2018
- Project Status: Active
- Research Area: Transportation Safety and Traffic Flow