Hybrid C-LSTM Neural Architecture for Real-Time Adaptive Control of Fixed-Wing AAVs in Variable Turbulence Conditions
IEEE Transactions on Automation Science and Engineering · 2026 · 23 · 7081-7099
DOI: 10.1109/tase.2026.3674571 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Automation Science and Engineering, 2026. Verfasst von Liguo Tan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 66 Literaturangaben. Gefördert von: Key Research and Development Program of Heilongjiang, National Natural Science Foundation of China, Key Research and Development Program of Zhejiang Province.
Abstract (Originalsprache)
Fixed-wing autonomous aerial vehicles (AAVs) performing turn maneuvers in turbulent environments face significant control challenges due to stochastic aerodynamic disturbances. This paper proposes a hybrid turbulence-aware control framework that integrates mathematical modeling, global optimization, and machine learning for adaptive autopilot gain tuning. A semi-invariant-based deterministic reduction of the stochastic lateral-directional model is developed, reducing the dimensionality from 44 to 7 equations while preserving control-relevant statistical characteristics. To determine optimal feedback gains, a Modified Survival of the Fittest Algorithm (MSoFA) is introduced, demonstrating improved global convergence reliability and achieving 5–15% better solution quality compared to conventional evolutionary methods. An offline-generated dataset of optimal gains across varying turbulence intensities and flight altitudes is then used to construct a clustered Long Short-Term Memory (C-LSTM) architecture that combines DBSCAN-based regime identification with recurrent neural approximation. The proposed neural controller achieves 98.3% gain prediction accuracy and reduces computation time by more than three orders of magnitude relative to direct optimization. Extensive numerical simulations confirm stable turn performance over altitudes ranging from 800 to 1500 m and turbulence scales from 150 to 470 m, with terminal yaw-rate deviations below 0.01%. The novelty of the proposed framework lies in the structured integration of 1) semi-invariant-based stochastic model reduction, 2) globally reliable evolutionary optimization for constrained gain tuning, and 3) regime-dependent recurrent neural approximation for real-time implementation. To the best of the authors’ knowledge, such an integrated turbulence-aware tuning framework for fixed-wing AAV autopilots has not been previously reported. Note to Practitioners—This paper is motivated by the difficulty of maintaining precise flight trajectories for fixed-wing Autonomous Aerial Vehicles (AAVs) in turbulent wind conditions, which is critical for applications ranging from agricultural monitoring to package delivery. Standard autopilots often lack the adaptability for sudden gusts, while sophisticated real-time optimization typically requires computing power unavailable on small, battery-powered onboard computers. This work addresses this trade-off by introducing a hybrid control framework that shifts the heavy computational burden offline, using a robust optimization method to generate a dataset of optimal control parameters that subsequently trains a specialized neural network for onboard use. The resulting system enables real-time adaptation, operating approximately 2000 times faster than traditional mathematical optimization and capable of running on standard embedded hardware with negligible latency. The primary contribution is a practical methodology for implementing sophisticated, turbulence-aware control on resource-constrained platforms without sacrificing accuracy. However, practitioners should note that the controller’s reliability is strictly limited to the operational envelopes, such as altitude ranges and turbulence intensities, defined in the training dataset; performance is not guaranteed outside these learned boundaries without retraining. Beyond aerial vehicles, this strategy of replacing computationally expensive optimization with fast neural approximation is applicable to other autonomous systems, such as underwater vehicles or ground robotics, that require rapid adaptive control in dynamic environments.
Förderung
- Key Research and Development Program of Heilongjiang
- National Natural Science Foundation of China
- Key Research and Development Program of Zhejiang Province
Zitieren
APA
Tan, L., Xiong, Y., Wen, G., Sun, K., & You, K. (2026). Hybrid C-LSTM Neural Architecture for Real-Time Adaptive Control of Fixed-Wing AAVs in Variable Turbulence Conditions. IEEE Transactions on Automation Science and Engineering, 23, 7081-7099. https://doi.org/10.1109/tase.2026.3674571
Vancouver
Tan L, Xiong Y, Wen G, Sun K, You K. Hybrid C-LSTM Neural Architecture for Real-Time Adaptive Control of Fixed-Wing AAVs in Variable Turbulence Conditions. IEEE Transactions on Automation Science and Engineering. 2026;23:7081-7099. doi:10.1109/tase.2026.3674571
BibTeX
@article{Tan2026,
author = {Tan, Liguo and Xiong, Yongcheng and Wen, Guanghui and Sun, Kangkang and You, Keyou},
title = {Hybrid C-LSTM Neural Architecture for Real-Time Adaptive Control of Fixed-Wing AAVs in Variable Turbulence Conditions},
journal = {IEEE Transactions on Automation Science and Engineering},
year = {2026},
volume = {23},
pages = {7081--7099},
doi = {10.1109/tase.2026.3674571}
}
EUVIMED