Abstract: The process to certify highly automated vehicles (AVs) has not yet been defined by any country in the world. Currently, companies test AVs on the public road, which is expensive and risky. I proposed the Accelerated Evaluation concept, which uses modified statistics of the surrounding vehicles and the Importance Sampling theory to reduce the evaluation time by several orders of magnitude, while ensuring the evaluation results are statistically accurate. The scenarios of the driving environment were taxonomized and modeled based-on 37 million miles of driving data collected by U-M. I developed mixed reality as a test tool on the proving ground to implement the approach and provide testing service to AV makers. I am also helping governments to establish testing protocols. My work has been supported by Ford, Toyota, Denso, Shanghai Motor, SFMotor, U.S. National Institute for Occupational Safety and Health, and China Automotive Technology Research Center. The vehicular knowledge (i.e., safety, energy, emissions, design) gained have the potential to be applied to many other types of intelligent physical systems.
Biosketch: Ding Zhao is a research scientist in the Department of Mechanical Engineering at the University of Michigan, Ann Arbor. He is also affiliated with UM Robotics Institute and Michigan Institute of Data Science. He graduated from Mechanical Engineering at UM in 2016 and did a one-year postdoc at UM Transportation Research Institute. His research is at the intersection of model-based control and model-free learning, with applications in autonomous driving, connected/smart cities, transportation energy efficiency in self-driving.
Date(s) - Apr 03, 2018
1:30 pm - 2:30 pm