| 作成者 |
|
|
|
|
|
| 本文言語 |
|
| 会議情報 |
|
| アクセス権 |
|
| 権利関係 |
|
|
|
| 関連DOI |
|
| 以下を参照 |
|
| 関連ISBN |
|
| 概要 |
This study proposes a deep reinforcement learning-based diversion assignment method to improve air traffic network resilience during unexpected airport disruptions. The accident at Haneda Airport on ...January 2, 2024, is used as a case study, in which 58 inbound flights are sequentially assigned to alternative airports or returned to their departure airports using a Deep Q-Network (DQN). The decision process considers operational constraints of candidate airports and the effects of diversion decisions on subsequent network operations. To evaluate the resulting recovery process more realistically, a resilience assessment method based on the temporal distribution of actual flight operations under normal conditions is introduced. The proposed framework enables passenger-oriented and resilience- oriented diversion strategies to be compared in terms of diversion-airport assignments, passenger cost, and network recovery time. The results show that the resilience-oriented strategy achieved the earliest complete network recovery, 132 min earlier than the actual operation and 188 min earlier than the passenger-oriented strategy, although at a higher estimated passenger cost.続きを見る
|
Mendeley出力
