Zhang Tianchen, Ding Yibo, Yue Xiaokui, Li Naying
School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China; Research & Development Institute of Northwestern Polytechnical University in Shenzhen, Sanhang Science &Technology Buliding, No. 45th, Gaoxin South 9th Road, Nanshan District, Shenzhen 518063, China; National Key Laboratory of Aerospace Flight Dynamics, Northwestern Polytechnical University, Xi'an 710072, China.
School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China; Beijing Electro-mechanical Engineering Institute, Beijing 100074, China.
ISA Trans. 2025 May;160:1-18. doi: 10.1016/j.isatra.2025.03.001. Epub 2025 Mar 8.
A data-driven adaptive terminal super-twisting prescribed performance controller (DASTPC) is designed for near-space vehicle (NSV) to satisfy transient and steady-state performance, and prevent scramjet choking. Firstly, a novel predetermined-time performance function is proposed to guarantee that tracking error can converge to a prescribed bound of small residual sets at the predetermined time. Compared with traditional performance functions, the predetermined-time performance function can achieve faster respond speed, realize more accurate convergence, and avoid overlarge initial value of actuators. Secondly, by combining the predetermined-time performance function with sliding mode control, a novel non-singular fast terminal sliding surface and an improved adaptive super-twisting reaching law are proposed to improve computational efficiency and accelerate convergent rate of system. The adaptive reaching law can avoid excessive gains and attenuate chattering by automatically tuning control gain. Thirdly, a deep recurrent neural network-based long-short term memory (LSTM) is employed to learn time-series historical flight dynamics data offline, so as to construct a data-driven LSTM training model. This data-driven model replaces nominal dynamics model of NSV in DASTPC, effectively suppressing model uncertainties. In addition, a homogeneous high-order sliding mode observer is utilized to compensate for external disturbances, avoiding excessive parameter estimation. Since boundary conditions of the predetermined-time performance function are fully satisfied, the DASTPC can effectively restrict amplitude of angle of attack, thus ensuring the intake condition of scramjet. Ultimately, to illustrate the superiority of DASTPC, several sets of simulations are performed on NSV subject to prescribed performance bound, external disturbances and parameter perturbations.
为了满足近空间飞行器（NSV）的瞬态和稳态性能，并防止超燃冲压发动机堵塞，设计了一种数据驱动的自适应终端超扭曲预设性能控制器（DASTPC）。首先，提出了一种新颖的预设时间性能函数，以确保跟踪误差能在预设时间收敛到小残差集的预设边界内。与传统性能函数相比，预设时间性能函数能够实现更快的响应速度，实现更精确的收敛，并避免执行器的初始值过大。其次，通过将预设时间性能函数与滑模控制相结合，提出了一种新颖的非奇异快速终端滑模面和一种改进的自适应超扭曲趋近律，以提高计算效率并加快系统的收敛速度。自适应趋近律可以避免增益过大，并通过自动调整控制增益来减弱抖振。第三，采用基于深度递归神经网络的长短期记忆（LSTM）离线学习时间序列历史飞行动力学数据，从而构建数据驱动的LSTM训练模型。该数据驱动模型取代了DASTPC中NSV的标称动力学模型，有效抑制了模型不确定性。此外，利用齐次高阶滑模观测器来补偿外部干扰，避免参数估计过多。由于完全满足了预设时间性能函数的边界条件，DASTPC能够有效地限制攻角的幅度，从而确保超燃冲压发动机的进气条件。最后，为了说明DASTPC的优越性，对受预设性能边界、外部干扰和参数摄动影响的NSV进行了几组仿真。
