Federated CNN-based sensor fault detection in unmanned aerial vehicles with WGAN-GP data augmentation

Shakibaeifard, Fatemeh; Eslaminejad, Mohammadreza; Peyravi, Farzad

doi:10.22075/ijnaa.2026.42239.5769

Federated CNN-based sensor fault detection in unmanned aerial vehicles with WGAN-GP data augmentation

Articles in Press

Document Type : Research Paper

Authors

Fatemeh Shakibaeifard ¹

Mohammadreza Eslaminejad ²

Farzad Peyravi ³

¹ Department of Computer Engineering, KI.C., Islamic Azad University, Kish, Iran

² Department of Computer Engineering, Zand Institute of Higher Education, Shiraz, Iran

³ Department of Computer Engineering, Marv.C., Islamic Azad University, Marvdasht, Iran

10.22075/ijnaa.2026.42239.5769

Abstract

Unmanned aerial vehicles (UAVs) depend on navigation sensors, including GPS, gyroscopes, accelerometers, magnetometers, barometers, and remote-control systems, for safe and autonomous flight. Sensor faults can cause trajectory deviation, flight instability, and mission failure. Existing fault detection approaches face challenges associated with class imbalance, distributed training, and multi-class fault discrimination. This study presents an integrated framework for seven-class UAV sensor status classification using the publicly available BASiC dataset. The framework combines a convolutional neural network with multi-scale feature extraction, residual blocks, and channel-wise attention with FedProx-based federated learning across multiple simulated clients representing distributed UAV nodes. Raw training data remain local to each client, while only model parameters are exchanged and aggregated. To address class imbalance, WGAN-GP generates synthetic samples for minority classes exclusively within the training set. Two scenarios are evaluated using the same held-out test set: federated CNN without augmentation and federated CNN with WGAN-GP augmentation. Flight-level data partitioning and training-only standardization and feature selection prevent test-set information leakage. The results show that federated CNN achieves 84.8\% accuracy, whereas WGAN-GP augmentation increases accuracy to 97.3\%, representing a 12.5-percentage-point improvement, with a comparable increase in macro-averaged F1-score. These findings demonstrate the potential of combining federated optimization and generative augmentation for UAV sensor fault classification in simulated decentralized training environments without transferring raw training data between clients.

Keywords

Unmanned Aerial Vehicles (UAVs)

Sensor Fault Detection

Convolutional Neural Networks (CNN)

Federated Learning

Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP)

Data Augmentation

Subjects

Machine Learning

[1] Z. Adaika, L.A. Al‑Haddad, W. Giernacki, A.A. Jaber, M. Boumehraz, M.N. Hamzah, and M.A. Flayyih, Fault detection and diagnosis methodologies for unmanned aerial vehicles: State‑of‑the‑art, J. Intell. Robot. Syst. 111 (2025), no. 2, 63.

[2] M.W. Ahmad and M.U. Akram, UAV sensor failures dataset: Biomisa arducopter sensory critique (BASiC), Data Brief 52 (2024), 110069.

[3] M.W. Ahmad, M.U. Akram, M.M. Mohsan, K. Saghar, R. Ahmad, and W.H. Butt, Transformer‑based sensor failure prediction and classification framework for UAVs, Expert Syst. Appl. 248 (2024), 123415.

[4] X. Chu, X. Zhou, Q. Bu, and Q. Miao, Sensor fault detection for UAVs using improved self‑attention LSTM network with similarity space mapping, IEEE Trans. Instrum. Meas. 73 (2024).

[5] M.H.M. Ghazali and W. Rahiman, A novel fault detection approach in UAV with adaptation of fuzzy logic and sensor fusion, IEEE/ASME Trans. Mechatronics 30 (2024), no. 1.

[6] I. Goodfellow, J. Pouget‑Abadie, M. Mirza, B. Xu, D. Warde‑Farley, S. Ozair, A. Courville, and Y. Bengio, Generative adversarial networks, Commun. ACM 63 (2020), no. 11, 139-144.

[7] A. Gretton, K.M. Borgwardt, M.J. Rasch, B. Scholkopf, and A. Smola, A kernel two‑sample test, J. Mach. Learn. Res. 13 (2012), no. 1, 723-773.

[8] I. Gulrajani, F. Ahmed, M. Arjovsky, V. Dumoulin, and A.C. Courville, Improved training of Wasserstein GANs, Adv. Neural Inf. Process. Syst. (NeurIPS) 30 (2017), 5767-5777.

[9] M.N. Hasan, S.U. Jan, and I. Koo, Wasserstein GAN‑based digital twin‑inspired model for early drift fault detection in wireless sensor networks, IEEE Sensors J. 23 (2023), no. 12, 13327-13339.

[10] K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 770-778.

[11] J. Hu, L. Shen, and G. Sun, Squeeze‑and‑excitation networks, Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2018, pp. 7132-7141.

[12] D.P. Kingma and J. Ba, Adam: A method for stochastic optimization, Int. Conf. Learn. Represent. (ICLR), 2015.

[13] R. Kohavi, A study of cross‑validation and bootstrap for accuracy estimation and model selection, Proc. 14th Int. Joint Conf. Artif. Intell. (IJCAI), 1995, pp. 1137-1143.

[14] A. Kumar, S. Wang, A.M. Shaikh, H. Bilal, B. Lu, and S. Song, Building on prior lightweight CNN model combined with LSTM‑AM framework to guide fault detection in fixed‑wing UAVs, Int. J. Mach. Learn. Cybern. 15 (2024), no. 9, 1-17.

[15] C. Li, K. Luo, L. Yang, S. Li, H. Wang, X. Zhang, and Z. Liao, A zero‑shot fault detection method for UAV sensors based on a novel CVAE‑GAN model, IEEE Sensors J. 24 (2024), no. 14.

[16] T. Li, A.K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V. Smith, Federated optimization in heterogeneous networks, Proc. Mach. Learn. Syst. (MLSys) 2 (2020), 429-450.

[17] I. Loshchilov and F. Hutter, SGDR: Stochastic gradient descent with warm restarts, Int. Conf. Learn. Represent. (ICLR), 2017.

[18] B. McMahan, E. Moore, D. Ramage, S. Hampson, and B.A. y Arcas, Communication‑efficient learning of deep networks from decentralized data, Proc. 20th Int. Conf. Artif. Intell. Stat. (AISTATS) (2017), 1273-1282.

[19] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, and J. Vanderplas, Scikit‑learn: Machine learning in Python, J. Mach. Learn. Res. 12 (2011), 2825-2830.

[20] V. Sadhu, K. Anjum, and D. Pompili, On‑board deep‑learning‑based unmanned aerial vehicle fault cause detection and classification via FPGAs, IEEE Trans. Robot. 39 (2023), no. 4, 3319-3331.

[21] C.B. Sahin, Quantum‑resilient federated learning for multi‑layer cyber anomaly detection in UAV systems, Sensors 26 (2026), no. 2, 509.