Academic Journal
Enhancing semantic and risk controllability in safety-critical scenario generation: An LLM-guided conditional diffusion method.
| Title: | Enhancing semantic and risk controllability in safety-critical scenario generation: An LLM-guided conditional diffusion method. |
|---|---|
| Authors: | Zhou H; School of Traffic and Transportation Engineering, Central South University, Changsha, China. Electronic address: hanchuzhou@csu.edu.cn., Li Y; School of Traffic and Transportation Engineering, Central South University, Changsha, China. Electronic address: yaqili19@163.com., Wang C; School of Traffic and Transportation Engineering, Central South University, Changsha, China; Shandong Hi-speed Group Innovation Research Institute, Jinan, China; State Key Lab of Intelligent Transportation System, Jinan, China. Electronic address: wangchengcheng@sdhsg.com., Chang F; School of Resources and Safety Engineering, Central South University, Changsha, China. Electronic address: 222023@csu.edu.cn., Xu P; School of Traffic and Transportation Engineering, Central South University, Changsha, China. Electronic address: pengpengxu@yeah.net., Li C; School of Traffic and Transportation Engineering, Central South University, Changsha, China. Electronic address: 254201017@csu.edu.cn. |
| Source: | Accident; analysis and prevention [Accid Anal Prev] 2026 Oct; Vol. 236, pp. 108648. Date of Electronic Publication: 2026 Jul 11. |
| Publication Type: | Journal Article; Validation Study |
| Language: | English |
| Journal Info: | Publisher: Pergamon Press Country of Publication: England NLM ID: 1254476 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1879-2057 (Electronic) Linking ISSN: 00014575 NLM ISO Abbreviation: Accid Anal Prev Subsets: MEDLINE |
| Imprint Name(s): | Publication: Oxford : Pergamon Press Original Publication: [New York, Pergamon Press] |
| MeSH Terms: | Accidents, Traffic*/prevention & control , Large Language Models* , Semantics* , Autonomous Vehicles* , Safety Management*, Natural Language Processing ; Computer Simulation ; Humans |
| Abstract: | Safety-critical scenario generation is essential for autonomous vehicle validation. However, existing methods often struggle to simultaneously achieve semantic controllability, risk-level controllability, and trajectory realism. To address this limitation, we propose a Large Language Model (LLM)-guided conditional diffusion framework for controllable safety-critical scenario generation. Specifically, we first use an LLM to convert natural-language descriptions into structured generation conditions. These conditions are then incorporated into a conditional latent diffusion model to guide scenario generation with controllable semantic and risk features. The generated candidate scenarios are further filtered to retain those with high semantic consistency. Experiments are conducted on the highD dataset for car-following and cut-in scenarios. The results show that the proposed method can generate realistic and risk-controllable scenarios. Compared with representative baseline methods, it achieves a better balance among criticality, realism, and controllability. Ablation experiments further demonstrate that LLM-based semantic conditioning improves both scenario-type and risk-level accuracy. The proposed method provides a flexible and controllable tool to generate safety-critical scenarios for autonomous vehicle testing and safety evaluation. (Copyright © 2026. Published by Elsevier Ltd.) |
| Competing Interests: | Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. |
| Contributed Indexing: | Keywords: Autonomous vehicles; Diffusion models; Large language models; Safety testing; Traffic safety |
| Entry Date(s): | Date Created: 20260711 Date Completed: 20260924 Latest Revision: 20260924 |
| Update Code: | 20260924 |
| DOI: | 10.1016/j.aap.2026.108648 |
| PMID: | 42435537 |
| Database: | MEDLINE |
| ISSN: | 1879-2057 |
|---|---|
| DOI: | 10.1016/j.aap.2026.108648 |