Generating Traffic Safety Test Scenarios for Automated Vehicles using a Big Data Technique

This study proposes a Big Data technique-based crash scene re-construction method in order to provide a systematic and effective way of developing critical traffic safety test scenarios for automated vehicles. It is widely understood that crashes occur through conflicts between road users and unsafe traffic conditions. Crash data are thus a useful resource to capture unsafe traffic conditions. Nevertheless, in practice many of the automated vehicle traffic test scenarios have been developed based on engineer’s judgment or aggregated crash statistics and studies, because crash data are too abundant to be comprehensively investigated. The proposed Big Data technique-based method provides the following innovations in developing automated vehicle test scenarios: abundant crash data can be investigated at once and in a short period of time using a text-weight analysis, which is one of the representative Big Data analytics; a comprehensive set of automated vehicle test scenarios can be developed by taking all significant words that frequently appear in crash descriptions into consideration; and the method can even be flexible in terms of capturing significant words by excluding or including specific words and setting a threshold of the word frequency. This proposed Big Data technique-based method is validated in comparison with the resulting test scenarios from a manual investigation of crash data, and it was found that 14 of a total of 18 scenarios correspond to the scenarios from manual investigation and the other four scenarios are additionally derived by the proposed approach. With these innovations in mind, the Big Data technique-based method proposed in this study is not only a systematic and practical approach but also leads to the development of a comprehensive set of automated vehicle test scenarios, by comprehensively taking into consideration significant words of crash descriptions, which should be carefully considered in developing automated vehicle test scenarios.

• Record URL:
  https://doi.org/10.1007/s12205-019-1287-4
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/12267988
• Supplemental Notes:
  • © Korean Society of Civil Engineers and Springer-Verlag Berlin Heidelberg 2019.
• Authors:
  • So, Jaehyun Jason
  • Park, Inseon
  • Wee, Jeongran
  • Park, Sangmin
  • Yun, Ilsoo
• Publication Date: 2019-6

Language

• English

Media Info

• Media Type: Digital/other
• Features: References;
• Pagination: pp 2702-2712
• Serial:
  • KSCE Journal of Civil Engineering
  • Volume: 23
  • Issue Number: 6
  • Publisher: Korean Society of Civil Engineers
  • ISSN: 1226-7988
  • EISSN: 1976-3808
  • Serial URL: http://link.springer.com/journal/12205

Subject/Index Terms

• TRT Terms: Crash analysis; Crash data; Data analysis; Data mining; Forecasting; Intelligent vehicles; Urban highways; Vehicle tests
• Uncontrolled Terms: Big data
• Subject Areas: Data and Information Technology; Highways; Safety and Human Factors; Vehicles and Equipment;

Filing Info

• Accession Number: 01709329
• Record Type: Publication
• Files: TRIS
• Created Date: Jun 27 2019 2:53PM