基于元学习与耦合状态相关黎卡提方程的变形飞行器变形与飞行协调控制
Coordinated Control of Deformation and Flight for Morphing Aircraft via Meta-Learning and Coupled State-Dependent Riccati Equations
AI总结:
本文提出基于元学习与耦合状态相关黎卡提方程的变形飞行器变形与飞行协调控制方法,通过域对抗不变元学习共享表示并求解微分博弈,仿真验证了有效性。
摘要:
本文利用元学习(ML)和耦合状态相关黎卡提方程(CSDREs)研究变形飞行器(MA)的变形与飞行协调控制问题。我们的方法基于两个主要观察:不同变形条件下MA的动态模型共享一个与变形条件无关的表示函数,且特定的变形条件部分位于一组线性系数中。为此,采用域对抗不变元学习(DAIML)利用离线飞行数据学习共享表示。基于学习到的表示函数,将MA的变形与飞行协调控制建模为非合作微分博弈。通过求解一对CSDREs,可以推导出状态相关反馈控制解。为此,扩展了Lyapunov迭代以获得CSDREs的正半定(定)镇定解,并提供了所提算法的收敛性证明。最后,进行仿真研究以验证所开发的协调博弈控制策略的有效性。
英文摘要:
In this paper, the coordinated control problem of deformation and flight for morphing aircraft (MA) is studied by using meta-learning (ML) and coupled state-dependent Riccati equations (CSDREs). Our method is built on two principal observations that dynamic models of MA under varying morphing conditions share a morphing condition independent representation function and that the specific morphing condition part lies in a set of linear coefficients. To that end, the domain adversarially invariant meta-learning (DAIML) is employed to learn the shared representation with offline flight data. Based on the learned representation function, the coordinated control of the deformation and flight for MA is formulated as a non-cooperative differential game. The state-dependent feedback control solutions can be derived by addressing a pair of CSDREs. For this purpose, Lyapunov iterations are extended to obtain the positive semidefinite (definite) stabilizing solutions of the CSDREs, and the convergence proof of the proposed algorithm is provided. Finally, a simulation study is carried out to validate the efficacy of the developed coordinated game control strategies.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。