LAST-V2X: Language-Driven Adaptive Scenario Synthesis and Testing for Communication-Induced Failures in V2X Safety Applications
Wu, K., Zhang, H., Gan, R., Li, P., Jiang, J., & Zhou, Y. (2026).
arXiv preprint, arXiv:2610.30979.
Summary
LAST-V2X turns natural-language descriptions into simulation scenarios for testing connected-vehicle safety applications. It combines traffic simulation with communication modeling to examine how delayed, missing, or incorrect messages affect driving decisions. Paired simulations compare degraded information with ground-truth information to isolate failures caused by communication. The framework uses simulation feedback to refine ineffective tests and produce reproducible failure cases. Evaluations in CARLA demonstrate scenario generation and improved failure discovery across multiple V2X safety applications.