Morphing-wing-structure control based on lift load monitoring by integrating optical fiber sensing and deep reinforcement learning

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K04850
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 24010:Aerospace engineering-related
• Research Institution: Japan Aerospace EXploration Agency
• Principal Investigator: Wada Daichi 国立研究開発法人宇宙航空研究開発機構, 航空技術部門, 研究開発員 (10770480)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 深層強化学習 / 可変翼 / 揚力同定 / 機械学習 / 光ファイバセンシング / 構造強度

Outline of Final Research Achievements

Focusing on the wings of unmanned aerial vehicles, this study has developed a technique to identify lift loads based on strain distributions measured by optical fiber sensors. In addition, based on the identified lifts loads and wind directions, a control technique has developed to reduce structural loads in real time. For the identification and control, neural networks were deployed, which were generated by machine learning and deep reinforcement learning, respectively. Furthermore, a morphing wing, which changed its bird-inspired wing shape, has designed and prototyped. The aerodynamic characteristics that varied in accordance with morphing modes were examined. These techniques have been tested in a wind tunnel and demonstrated their feasibility and effectiveness.

Free Research Field

航空宇宙工学

Academic Significance and Societal Importance of the Research Achievements

構造負荷低減という、より高次な構造運用目的に対して深層強化学習を適用する好例を示せた。とりわけ「構造状態をセンシングし、飛行環境を認識し、それに合わせて制御する」という体系的なシステムとして技術統合しており、知能的な構造運用の技術体系を提案できた。可変翼によるより豊かな空力表現・活用も含めて、風洞試験により実証できたことで、実用性のある工学的知見となった。