基于深度强化学习的变体飞行器切换抗扰动控制

Morphing aircraft can adaptively alter its aerodynamic configuration according to the environment and the requirement of missions. The key purpose of its control system is to improve the capability of intelligent morphing and stable flight, but the flight control methods for traditional aircraft become less effective. Therefore, this project first combines the advantages of deep learning in sense ability and reinforcement leaning in decision-making to investigate the intelligent morphing strategy. Then, the morphing aircraft is modeled by nonlinear switched systems. The uncertainties of the aircraft model and the external/internal disturbances are handled as total disturbance to the aircraft. Then, active anti-disturbance control method could be applied to the morphing aircraft nonlinear switched systems. This project also develops real-time hardware-in-the-loop simulation platform for morphing aircraft to validate the effectiveness of the proposed methods, and the feasibility for engineering application. This project would provide theoretical methods and engineering experiences for control system design of morphing aircraft in our country.

变体飞行器能根据飞行环境变化和飞行任务需求自适应改变外形，其控制系统设计的关键是提高飞行器智能自主变形和稳定飞行能力，传统飞行器控制系统设计方法已不能满足变体飞行器的需求。为此，本项目首先结合深度学习的感知能力优势和强化学习的决策能力优势，研究基于深度强化学习的变体飞行器智能自主变形策略；在此基础上，建立变体飞行器非线性切换系统模型，将变体飞行器模型中的各种不确定因素和内外部干扰统一视为对飞行器稳定控制的总扰动，研究面向变体飞行器的非线性切换系统主动抗扰动控制方法；最后，搭建变体飞行器半实物实时仿真平台，验证所提理论方法的有效性和工程推广的可行性。本项目的预期研究成果可为我国变体飞行器控制系统的研制奠定良好的理论基础和工程依据。

变体飞行器可以根据飞行环境变化和飞行任务需求自适应改变外形，相比传统飞行器具有能够通过主动变形改善气动特性、降低飞行能耗、增强控制能力等诸多优势。针对变体飞行器在复杂多变飞行环境与飞行任务下的自主变形和稳定飞行控制问题，本项目在深入分析变体飞行器的外形和姿态控制机理的基础上，建立了变体飞行器多刚体动力学模型与面向控制的非线性切换系统模型，设计了基于深度强化学习的变体飞行器自主变形策略，提出了变体飞行器非线性切换系统主动抗扰动控制方法，编制了基于变体飞行器动力学建模与控制仿真验证软件，验证了所提理论与方法的有效性以及工程推广的可行性。本项目为变体飞行器控制系统提供一套高效可行的设计、分析与实现方法，为有效利用“扰动观测+补偿” 设计框架和深度强化学习方法解决变体飞行器的动力学建模与控制问题奠定了良好的理论基础和工程依据。