Research
Projects
Development of a Large Language Model (LLM)-Driven Automated Test Scenario Generation Algorithm Based on Autonomous Vehicle Accident Records
자율주행 사고기록 기반 대형언어모델(LLM) 활용 자율차 테스트 시나리오 자동 생성 알고리즘 개발
Client: Korea Transportation Safety Authority
Lead Institution: Ajou University
Researchers: Gyungtaek Oh, Joonho Park, Jaehyun (Jason) So
Role of Ajou University
1. Autonomous driving accident data cleaning and structuring
2. LLM-based automatic classification of accident types and driving modes
3. Scenario similarity-based validation and suitability assessment
Project Description
- Automatically converting unstructured textual accident data into a form suitable for analysis and scenario generation through LLM-based structuring and standardization, securing a foundation for automatically updating and advancing the scenario catalog using continuously accumulated accident data, and realizing a data-based evaluation environment in which test scenarios are continuously updated as accident data accumulate
