Failure-Scenario Maker for Autonomous Driving Vehicle using Adversarial Multi-agent Reinforcement Learning

The author proposes a method to create failure-scenarios of an autonomous driving vehicle by training other surrounding vehicles by means of multiagent adversarial reinforcement learning. Failure in driving environments might lead to catastrophic results. Hence, when developing a software of autonomous driving cars, one must find as many failure-cases as possible and then improve the software. However, as the software becomes complicated, it is hard to find failure-scenarios that are useful for the software improvement. Hence, the author proposes a framework to create various failure-scenarios of an autonomous car by training other car(s) via reinforcement learning. He demonstrates the effectiveness of his proposed method with two kinds of experiments.

• Record URL:
  https://doi.org/10.11351/jsaeronbun.51.950
• Record URL:
  https://www.jstage.jst.go.jp/article/jsaeronbun/51/5/51_20204474/_pdf
• 
• Record URL:
  https://scholar.google.com/scholar_lookup?title=Failure-Scenario%20Maker%20for%20Autonomous%20Driving%20Vehicle%20using%20Adversarial%20Multi-agent%20Reinforcement%20Learning&author=A.%20Wachi&publication_year=2020
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/02878321
• Authors:
  • Wachi, Akifumi
• Publication Date: 2020-9

Language

• English
• Japanese

Media Info

• Media Type: Digital/other
• Features: Figures; Photos; References; Tables;
• Pagination: pp 950-955
• Serial:
  • Transactions of Society of Automotive Engineers of Japan
  • Volume: 51
  • Issue Number: 5
  • Publisher: Society of Automotive Engineers of Japan
  • ISSN: 0287-8321
  • EISSN: 1883-0811
  • Serial URL: https://www.jstage.jst.go.jp/browse/jsaeronbun
• Publication flags:

  Open Access (libre)

Subject/Index Terms

• TRT Terms: Autonomous vehicles; Crash avoidance systems; Driving simulators; Fail safe systems; Machine learning; Multi-agent systems; Software; Vehicle safety
• Subject Areas: Data and Information Technology; Highways; Safety and Human Factors; Vehicles and Equipment;

Filing Info

• Accession Number: 01756456
• Record Type: Publication
• Source Agency: Japan Science and Technology Agency (JST)
• Files: TRIS, JSTAGE
• Created Date: Oct 29 2020 9:25AM