The paper introduces Probabilistic Lane Graphs (PLGs) to generate realistic corner cases for AV safety assessment. By learning from spatio-temporal traffic data, the PLG model allows for explainable and human-understandable corner case scenarios. Reinforcement learning techniques are used to modify policies and generate complex yet explainable scenarios. The methodology is tested on real-world datasets like NGSIM, showcasing a significant improvement in corner case generation rates compared to existing methods.
To Another Language
from source content
arxiv.org
Дополнительные вопросы