Empowering the Edge Intelligence by Air-Ground Integrated Federated Learning in 6G Networks

Yuben Qu, Chao Dong, Jianchao Zheng, Qihui Wu, Yun Shen, Fan Wu, Alagan Anpalagan

Abstract

Ubiquitous intelligence has been widely recognized as a critical vision of the future sixth generation (6G) networks, which implies the intelligence over the whole network from the core to the edge including end devices. Nevertheless, fulfilling such vision, particularly the intelligence at the edge, is extremely challenging, due to the limited resources of edge devices as well as the ubiquitous coverage envisioned by 6G. To empower the edge intelligence, in this article, we propose a novel framework called AGIFL (Air-Ground Integrated Federated Learning), which organically integrates air-ground integrated networks and federated learning (FL). In the AGIFL, leveraging the flexible on-demand 3D deployment of aerial nodes such as unmanned aerial vehicles (UAVs), all the nodes can collaboratively train an effective learning model by FL. We also conduct a case study to evaluate the effect of two different deployment schemes of the UAV over the learning and network performance. Last but not the least, we highlight several technical challenges and future research directions in the AGIFL.

Introduction

The evolution toward 6G networks envisions "ubiquitous intelligence," where artificial intelligence is seamlessly integrated from the network core to the extreme edge. However, realizing this vision faces two primary obstacles: the severe resource constraints of edge devices (e.g., limited battery and compute) and the difficulty of providing continuous coverage in remote or highly dynamic environments.

While Federated Learning (FL) offers a privacy-preserving way to train models across distributed devices, its application in Air-Ground Integrated Networks (AGINs) introduces unique challenges. The mobility and energy limitations of aerial nodes, such as Unmanned Aerial Vehicles (UAVs), necessitate sophisticated coordination to ensure efficient model training.

The AGIFL Framework

The proposed Air-Ground Integrated Federated Learning (AGIFL) framework leverages the mobility of aerial nodes to provide on-demand intelligence. The framework categorizes interactions into four distinct scenarios based on the roles of aerial and terrestrial nodes:

  • Air-to-Air (A2A): A swarm of UAVs performs collaborative training with one aerial node acting as the parameter server.
  • Ground-to-Air (G2A): A UAV functions as a mobile parameter server for ground-based IoT devices, bridging coverage gaps.
  • Air-to-Ground (A2G): Terrestrial base stations serve as parameter servers for resource-constrained aerial participants.
  • Mixed: A flexible configuration where both aerial and terrestrial nodes can act as either participants or servers.
The four interaction scenarios in the AGIFL framework: (a) A2A, (b) G2A, (c) A2G, and (d) Mixed federated learning.

Case Study and Performance Analysis

To evaluate the framework, the authors conducted a case study in a G2A scenario using the MNIST dataset. They compared a random UAV deployment against an optimized Min_SumDist scheme, which uses convex optimization to find the ideal hovering coordinate to minimize total communication distance.

The results demonstrate that physical deployment is a critical "hyperparameter" in AGIFL. Optimized positioning significantly reduces communication energy and latency. Because the UAV can hover more efficiently, it can sustain more training rounds within a fixed energy budget, leading to higher model accuracy. For instance, at an energy budget of 220 Joules, the optimized scheme achieved approximately 93% accuracy, whereas the random scheme reached only 88%.

Testing accuracy versus UAV energy budget, comparing optimized (Min_SumDist) and random deployment strategies.

Conclusion and Future Directions

The AGIFL framework provides a promising path toward ubiquitous 6G intelligence by treating aerial nodes as dynamic components of the compute stack. Future research will focus on handling non-IID data distributions, improving robustness against unreliable wireless channels, and enhancing security through decentralized peer-to-peer models and blockchain integration.

Summary updated

Abstract

Ubiquitous intelligence has been widely recognized as a critical vision of the future sixth generation (6G) networks, which implies the intelligence over the whole network from the core to the edge including end devices. Nevertheless, fulfilling such vision, particularly the intelligence at the edge, is extremely challenging, due to the limited resources of edge devices as well as the ubiquitous coverage envisioned by 6G. To empower the edge intelligence, in this article, we propose a novel framework called AGIFL (Air-Ground Integrated Federated Learning), which organically integrates air-ground integrated networks and federated learning (FL). In the AGIFL, leveraging the flexible on-demand 3D deployment of aerial nodes such as unmanned aerial vehicles (UAVs), all the nodes can collaboratively train an effective learning model by FL. We also conduct a case study to evaluate the effect of two different deployment schemes of the UAV over the learning and network performance. Last but not the least, we highlight several technical challenges and future research directions in the AGIFL.
BibTeX
@misc{qu2020empowering,
  author = {Yuben Qu and Chao Dong and Jianchao Zheng and Qihui Wu and Yun Shen and Fan Wu and Alagan Anpalagan},
  title = {{Empowering the Edge Intelligence by Air-Ground Integrated Federated Learning in 6G Networks}},
  year = {2020},
  eprint = {2007.13054},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2007.13054},
}
Figure