Key points are not available for this paper at this time.
Transportation systems face increasingly complex disruptions that challenge traditional safety and resilience frameworks reliant on historical stationarity. Generative Artificial Intelligence (GAI)—including generative adversarial networks, diffusion models, and large language models—offers novel capabilities for rare-event simulation, multimodal data augmentation, and proactive scenario generation. This paper conducts a systematic review of 170 peer-reviewed studies published between 2019 and 2025, integrating fragmented findings into a lifecycle perspective spanning accident prevention, accident prediction, accident response and recovery. Our analysis demonstrates that GAI enables significant advances in accident prevention through enhanced traffic flow and behavior prediction, improves accident forecasting via anomaly and collision detection, and supports real-time emergency response and post-disruption recovery through multimodal reasoning and decision support. We also identify critical challenges in interpretability, data quality, domain adaptation, and ethical governance, which constrain GAI’s safe deployment in transportation systems. Building on these insights, we outline future research directions emphasizing human–AI collaboration, bias mitigation, multimodal integration, and governance mechanisms for trustworthy and resilient applications. This review contributes to the operations and transportation management literature by providing comprehensive, lifecycle-oriented synthesis of GAI in transportation safety and resilience, and by highlighting pathways for aligning advanced AI technologies with the design of next-generation resilient infrastructure systems.
Liu et al. (Wed,) studied this question.