Shuo Feng

The severe inefficiency of safety validation for vehicles has become a critical bottleneck in the development and deployment of AV systems. It is widely recognized that validating the safety of an autonomous vehicle with high-confidence may require tens of billions of miles of testing in natural driving environments.

To tackle this industry challenge, Shuo Feng focused on the underlying scientific issue (rare-event estimation problem in a high-dimensional space) and proposed a new approach called " Intelligent Testing with Continuous Spatio-Temporal Environment Modeling".

Shuo established the theoretical framework and methodology of "Equivalent Accelerated Testing for Autonomous Vehicle" to address the inefficiency of testing and overcome the limitations of scenario-based testing. This framework enables intelligent generation of testing environments and significantly enhances the testing capabilities across broad temporal and spatial scales. Utilizing an "AI Against AI" strategy, it significantly accelerates the testing process by 3 to 5 orders of magnitude.

The groundbreaking paper was featured as the cover article in Nature in March 2023. It marks the first time a paper on autonomous driving has been published in Nature.