Research Article
Federated Learning-Based Channel Estimation in Vehicular Networks
@INPROCEEDINGS{10.1007/978-3-031-96146-5_7, author={Cao Ding and Ivan Wang-Hei Ho}, title={Federated Learning-Based Channel Estimation in Vehicular Networks}, proceedings={Smart Grid and Innovative Frontiers in Telecommunications. 8th EAI International Conference, EAI SmartGIFT 2024a, Santa Clara, United States, March 23-24, 2024, Proceedings}, proceedings_a={SMARTGIFT}, year={2026}, month={9}, keywords={Federated Learning Channel Estimation Vehicular Networks}, doi={10.1007/978-3-031-96146-5_7} }- Cao Ding
Ivan Wang-Hei Ho
Year: 2026
Federated Learning-Based Channel Estimation in Vehicular Networks
SMARTGIFT
Springer
DOI: 10.1007/978-3-031-96146-5_7
Abstract
Most studies on channel estimation make assumptions on statistical channel models, such as the Rayleigh fading model, but these assumptions are impractical. On one hand, the high-speed mobility of vehicles generates carrier frequency offset (CFO) and Doppler shift, which leads to inter-carrier interference (ICI). On the other hand, buildings in the connected vehicle environment cause propagation loss to the communication link, which is known as the urban canyon effect. To solve these issues, this paper exploits ray tracing technology to model the propagation loss and proposes a channel estimator based on federated learning and deep neural network (FL-DNN) techniques. The proposed FL-DNN estimator divides multiple geographical regions through K-D tree for local model training and uses trimmed mean averaging to aggregate the global model. Our results show that the channel estimation performance and data recovery accuracy of FL-DNN outperform traditional and learning-based channel estimators. Our results show that FL-DNN outperforms traditional and learning-based channel estimators by an average of 63.5% and 57.2% in channel estimation and data recovery accuracy, respectively.