本研究は広域洪水後の道路復旧作業の最適化において、人流を考慮した深層強化学習モデリングの開発を行ったものです。とくに2018年の西日本豪雨を対象に、2.32*10^52のケースから約3時間で最適な道路復旧戦略を抽出できるようになりました。これにより道路管理者は災害直後に、適切な時間で、より効率的な道路復旧計画ができる事が期待されます。
In this study, considering the people flow, we developed a deep reinforcement learning model to optimize the road restoration work after the widespread flooding. In the case of the Heavy Rain Event of July 2018 in Japan, we were able to extract the optimal road restoration strategy from 2.32*10^52 cases in about 3 hours. This is expected to enable road managers to make more efficient road restoration plans immediately after the disaster within a reasonable time.
【論文情報】
Soo-hyun Joo, Yoshiki Ogawa, Yoshihide Sekimoto, Road-reconstruction after multi-locational flooding in multi-agent deep RL with the consideration of human mobility – Case study: Western Japan flooding in 2018 -, International Journal of Disaster Risk Reduction, Elsevier, 70, Jan. 2022, 102780, DOI (Impact factor in 2021: 4.320)
図：本研究で開発したモデルの概要
Figure: Framework of proposed model.
