Promotion of Proactive Behavioral Change to Prevent Human Error (Part 5) Development of a Fundamental Technology for Integrating Attention Targets Extracted via LLM-based Traffic Context Understanding with Driver Gaze Targets
ヒューマンエラー防止のための主体的行動変容の促進 (第5報)大規模言語モデルによる交通環境文脈理解に基づく注意対象抽出とドライバ視認行動の統合のための基盤技術開発
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.):¥1,100 Members (tax incl.):¥880
- Publication code
- 20266357
- Paper/Info type
- Proceedings (Autumn)
No.166-26
- Pages
- 1-2(Total 2 p)
- Date of publication
- Oct 2026
- Publisher
- JSAE
- Language
- Japanese
- Event
- 2026 JSAE Annual Congress (Autumn)
Detailed Information
| Author(J) | 1) 山﨑 壮馬, 2) 山崎 彬人, 3) 吉武 宏, 4) 小竹 元基 |
|---|---|
| Author(E) | 1) Soma Yamasaki, 2) Akito Yamasaki, 3) Hiroshi Yoshitake, 4) Motoki Shino |
| Affiliation(J) | 1) 名城大学, 2) 名城大学, 3) 東京科学大学, 4) 東京科学大学 |
| Affiliation(E) | 1) Meijo University, 2) Meijo University, 3) Institute of Science Tokyo, 4) Institute of Science Tokyo |
| Abstract(J) | 従来の運転行動分析では,ドライバが「何を見たか」は評価できても,「何を見るべきだったか」という交通環境側の要求を扱うことは困難であった.本研究では,大規模言語モデルにより道路環境から「注意対象」を抽出・構造化し,ドライバの「視認対象」と統合することで,確認行動の妥当性を定量評価する手法を提案する. Translation |
| Abstract(E) | Conventional driving behavior analysis can evaluate "what the driver looked at," but it has been difficult to address the requirements imposed by the traffic environment, namely "what the driver should have looked at." In this study, we propose a method that extracts and structures "attention targets" from road environment images using a large language model, and integrates them with the driver's "gaze targets" to quantitatively evaluate the appropriateness of confirmation behavior. |