Development of Real-Time Traffic State Estimation Methods for Next-Generation Traffic Control

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K14266
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 22050:Civil engineering plan and transportation engineering-related
• Research Institution: The University of Tokyo
• Principal Investigator: YASUDA SHOHEI 東京大学, 大学院工学系研究科(工学部), 助教 (00899247)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 交通流理論 / 深層学習 / ネットワーク表現 / 空間統計学

Outline of Final Research Achievements

In this study, we developed a method for real-time traffic state estimation on large-scale road networks, aiming at next-generation traffic control. Specifically, we focused on dynamic network representations using spatial statistics and the development of traffic state estimation methods utilizing traffic flow theory and deep learning. By developing a methodology for dynamic network representation based on vehicle trajectory data tailored to specific purposes, we demonstrated the potential to significantly reduce computational complexity while ensuring estimation accuracy. Furthermore, by combining traffic flow theory and deep learning, we developed an estimation method capable of efficiently handling computations involving a vast number of parameters for the network, while physically describing the dynamics of congestion formation and dissipation.

Free Research Field

交通工学

Academic Significance and Societal Importance of the Research Achievements

本研究成果の学術的意義は，大規模なネットワークを対象とした交通状態推定において，その解析空間である道路ネットワークを観測データ自体から動的に生成するという新たなアプローチにより，推定精度を担保しつつ計算量を大幅に削減できる可能性を示した点である．社会的意義として，計算量の削減による高速な推定や渋滞の延伸解消を物理学的な観点を有しつつ記述可能な方法論の実現は，正確かつリアルタイムな交通状態推定を可能とし，災害時の効率的な避難誘導や自動運転等の新技術の統合的な制御等の次世代管制システムの実現に寄与する．