Exploring master scenarios for autonomous driving tests from police-reported historical crash data using an adaptive search sampling framework

Published in Accident Analysis & Prevention, 2024

Recommended citation: Li, Y., Yang, Z., Jin, J., & Wu, D. (2024). Exploring master scenarios for autonomous driving tests from police-reported historical crash data using an adaptive search sampling framework. Accident Analysis & Prevention, 205, 107688. https://www.sciencedirect.com/science/article/pii/S0001457524002331

’– Developed a novel approach using police-reported historical crash data to explore fundamental functional scenarios crucial for testing autonomous driving systems.

– Introduced the concept of master scenario samples as representative subsets of historical crash data. Using deep generative models to synthesize diverse derived scenarios samples.

– Proposed an adaptive search sampling method that iteratively refines sampling strategies based on the Similarity Score, ensuring comprehensive coverage of crash scenarios.

– Established a surrogate model (SM) to expedite the process of scenario samples generation and evaluation, addressing the computational challenges associated with training deep generative models.’

Download paper here

Recommended citation: Li, Y., Yang, Z., Jin, J., & Wu, D. (2024). Exploring master scenarios for autonomous driving tests from police-reported historical crash data using an adaptive search sampling framework. Accident Analysis & Prevention, 205, 107688.