Collaborative information processing platform for vehicular IoT based on federated learning

2023 Fiscal Year Final Research Report

• Project/Area Number: 21H03424
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Review Section: Basic Section 60060:Information network-related
• Research Institution: The University of Electro-Communications
• Principal Investigator: Wu Celimuge 電気通信大学, 大学院情報理工学研究科, 教授 (90596230)
• Co-Investigator(Kenkyū-buntansha): 村瀬 勉 名古屋大学, 情報基盤センター, 教授 (10530941) ; 李 鵬 (李鵬) 会津大学, コンピュータ理工学部, 上級准教授 (30735915) ; 計 宇生 国立情報学研究所, アーキテクチャ科学研究系, 教授 (80225333)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 車両IoT / フェデレーテッド・ラーニング / 協調型情報処理基盤技術

Outline of Final Research Achievements

In order to realize cooperative autonomous driving, etc., a highly efficient and high-performance vehicle IoT information processing infrastructure technology is required. In this study, we proposed a vehicle-to-vehicle cooperation and self-evolving control method based on the federated learning (FL) method that can handle complex and diverse dynamic environments. In order to improve the performance of existing FL, we conducted research on the following four technologies: (1) deep reinforcement learning-based FL technology, (2) pre-learning technology using fuzzy logic, and (3) learning client selection and model aggregation technology, (4) FL technology using blockchain in an autonomous decentralized environment. We conducted computer simulations and real-world experiments to evaluate the proposed approach.

Free Research Field

情報ネットワーク

Academic Significance and Societal Importance of the Research Achievements

本研究では，深層強化学習とFLの組み合わせによる通信・計算・ストレージ資源の共同最適化を実現し，車両間協調で各車両の行動改善を行う手法に関する議論を行った．FLを用いた車両IoT情報処理基盤は独創的な研究であり，車両IoTにおける協調知能の実現にとって重要な一歩である．本研究で提案した手法は，高効率かつ高性能な車両IoT情報処理基盤を実現するための重要な技術であり，今後は，協調型自動運転などの高度な連携協調システムにおける活用が期待される．