
Shuo Feng
Assistant Research Scientist
Location
University of Michigan Transportation Research Institute, 2901 Baxter Rd. Ann Arbor, MI 48109
Room 104
Primary Website
Education
- Ph.D., Control Science and Engineering, Tsinghua University, 2019
- Visiting Scholar, Civil and Environmental Engineering, University of Michigan, 2017-2019
- B.S. Automation, Tsinghua University, 2014
Research Interests
- Connected and Automated Vehicle Testing and Evaluation
- Cooperative Automation
- Traffic Behavior Modeling
Biography
Dr. Shuo Feng’s research work has been focused on the testing and evaluation of connected and automated vehicles. He has published 20 peer-reviewed journal papers, with 12 of them being the leading author. Dr. Feng’s dissertation is titled “Testing Scenario Library Generation for Connected and Automated Vehicles”. The topic is timely and important as the development of connected and automated vehicles (CAVs) has attracted significant attention from a wide spectrum of stakeholders, and yet there are no consensus nor standard procedures on how to test and evaluate CAVs. The theory and methods to support such testing and evaluation are significantly lacking. The topic is also challenging because it is interdisciplinary in nature and suffers from interdisciplinary difficulties. In his dissertation, Dr. Feng made the methodological breakthrough and developed a new framework for testing scenario generation, which can dramatically reduce the number of CAV tests without loss of evaluation unbiasedness. The breakthrough is enabled by his development of a new scenario evaluation metric, scenario criticality, to ensure that the selected scenarios are both naturalistic and adversarial, and the application of importance sampling theory to accelerate the evaluation process. The proposed methods have been implemented in the Mcity testing facility at the University of Michigan. Testing results at Mcity demonstrated that the proposed methods can accelerate the CAV evaluation process by hundreds or even thousands of times with the same accuracy. Based on this work, Dr. Feng won the IEEE Intelligent Transportation Systems Society’s Best Ph.D. Dissertation Award in 2020, which is given annually for the best dissertation in the field of intelligent transportation systems.
During his postdoctoral period, Dr. Feng significantly extended the testing scenario library generation framework for complex driving situations that involve various maneuvers of multiple traffic participants for an extended period. By training the background vehicles to learn when to execute what adversarial maneuver, the resulting test environment, which is called the naturalistic and adversarial driving environment (NADE), can continuously generate testing scenarios for a testing CAV in a highway driving environment. Since the base input for the scenario generation is from naturalistic driving data, the testing miles in test tracks can be converted approximately into equivalent mileage with on-road naturalistic driving environments. Because the challenging scenarios significantly increase the exposure of safety-critical cases, a testing mile in test tracks could be equivalent to hundreds or even thousands of driving miles on public roads. The paper from this research was published in Nature Communications, one of the most impactful journals covering all fields of natural sciences. The method is being implemented in both the Mcity test facility and American Center for Mobility (ACM), which is one of the leading test tracks in the world.
Awards
- Best Dissertation Award, IEEE Intelligent Transportation System Society (ITSS) (9/2020)
- Outstanding Graduates, Tsinghua University, China (7/2020)
- Best Dissertation Award, Tsinghua University, China (6/2020)
- National Scholarship, China (11/2017)
- Outstanding Undergraduates, Tsinghua University, China (7/2014)
- Outstanding Undergraduates, Beijing Province, China (6/2014)
ORCID: https://orcid.org/0000-0002-2117-4427
Website: https://www.fsuo.tech