Safety Assessment for Autonomous Vehicles - Henry Liu

Earning public trust in highly automated vehicles, and greater understanding of how they operate, is paramount as automakers and their suppliers push to put cars and trucks on the road that can safely operate without a driver. While complex automated vehicle safety testing programs are being implemented across industry, they are generally proprietary, and their structure remains hidden from public view. To remedy this, we developed the Mcity Safety Assessment Program, a two-part protocol for testing the behavior competence of automated vehicles before their widespread use on public roads. The first part of the assessment is a “Driver’s License Test” that measures the basic behavioral competency of an automated vehicle through random scenario generation. The second part is a “Driving Intelligence Test” that challenges AI-based algorithms with a diverse set of scenarios representing those that most often result in crashes, injuries and fatalities. Mcity believes the Mcity Safety Assessment Program could serve as the blueprint for a publicly inspectable behavioral safety framework, helping industry bring automated vehicle technology to market in a manner that truly benefits society. In this talk, we will also highlight Mcity 2.0, a facility funded by the National Science Foundation, that aims to build a digital infrastructure providing researchers remote access to the Mcity mixed reality testing environment for highly automated vehicles.