Blockchain-enabled Clustered and Scalable Federated Learning (BCS-FL) Framework in UAV Networks

TL;DR

This work tackles privacy, scalability, and reliability challenges in UAV-network FL by introducing BCS-FL, which partitions UAVs into clusters and uses blockchain-enabled smart contracts to manage registration, clustering, and aggregation. The global model is obtained through inter-cluster aggregation strategies, notably FCA and $k$-HA, with FCA computing $reve{l}^d=rac{1}{Q}\sum_{q\, ext{in}\,oldsymbol{mathcal{Q}}}l_q^d$ and $k$-HA sharing updates across CHs within $k$ hops; local updates follow FedAvg weighting with $oldsymbol{C3_u}=|oldsymbol{\mathcal{R}_u}|/|oldsymbol{\mathcal{R}_q}|$. Numerical results on MNIST and CIFAR-10 (IID and non-IID) show convergence and reveal trade-offs between accuracy and communication overhead, with FCA achieving near-centralized performance and 1HA offering lower overhead at the cost of slower diffusion. Overall, BCS-FL demonstrates a path toward scalable, privacy-preserving FL in UAV swarms, balancing collaboration efficiency and resource constraints in large, dynamic networks.

Abstract

Privacy, scalability, and reliability are significant challenges in unmanned aerial vehicle (UAV) networks as distributed systems, especially when employing machine learning (ML) technologies with substantial data exchange. Recently, the application of federated learning (FL) to UAV networks has improved collaboration, privacy, resilience, and adaptability, making it a promising framework for UAV applications. However, implementing FL for UAV networks introduces drawbacks such as communication overhead, synchronization issues, scalability limitations, and resource constraints. To address these challenges, this paper presents the Blockchain-enabled Clustered and Scalable Federated Learning (BCS-FL) framework for UAV networks. This improves the decentralization, coordination, scalability, and efficiency of FL in large-scale UAV networks. The framework partitions UAV networks into separate clusters, coordinated by cluster head UAVs (CHs), to establish a connected graph. Clustering enables efficient coordination of updates to the ML model. Additionally, hybrid inter-cluster and intra-cluster model aggregation schemes generate the global model after each training round, improving collaboration and knowledge sharing among clusters. The numerical findings illustrate the achievement of convergence while also emphasizing the trade-offs between the effectiveness of training and communication efficiency.

Figures (6)

• Figure 1: The BCS-FL Framework.
• Figure 3: Performance of (a) accuracy and (b) loss CIFAR-10 dataset.
• Figure 4: The BCS-FL Performance of accuracy.
• Figure 5: Performance on BCS-FL (a) accuracy and (b) loss MNIST dataset.
• Figure 6: Influence of $k$ on BCS-FL (a) accuracy and (b) loss.