Associate Research Scientist, Engineering Systems, University of Michigan Transportation Research Institute
Adjunct Associate Professor of Industrial and Operations Engineering, College of Engineering
This project examined approaches to evaluate the safety of automated vehicles (AVs) that reduce the test duration from a few months to a few days without sacrificing accuracy, and identified possible failure modes on perception errors and mis-operations. Tests were developed in two layers: scenarios and variations. Test scenarios were extracted from naturalistic driving data, crash data, and expert knowledge. For each scenario, test variations were developed based on naturalistic driving data collected in the Ann Arbor Connected project (Pillar 1). Statistical tools were developed, based on the importance sampling approach, to measure safety impacts efficiently by reducing the safe and boring events and emphasized the safety-critical conditions.
(1) Test scenarios and variations to evaluate the safety of AVs, and (2) estimates of safety benefit/risk to society after AVs are released and deployed in Ann Arbor.