Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficient UAV Swarm-Based Multi-Task Federated Learning with Dynamic Task Knowledge Sharing

Yubo Yang, Tao Yang, Xiaofeng Wu, Ziyu Guo, Bo Hu

TL;DR

The paper tackles multi-task federated learning for UAV swarms under strict energy and bandwidth limits, introducing a dynamic task attention mechanism and a task affinity metric to capture time-varying task importance and inter-task correlations. It develops a Lyapunov-based two-layer optimization to jointly optimize UAV transmission power, computation frequency, bandwidth allocation, and UAV–EV associations, with closed-form inner solutions and a two-stage outer association algorithm. The approach yields an O(\sqrt{V}, 1/V) trade-off between UAV energy consumption and multi-task performance and demonstrates improved convergence and generalization through targeted knowledge sharing across related tasks. Empirical results on MNIST, FMNIST, and FLAME-derived UAV datasets confirm faster convergence, higher accuracy, and better resource efficiency compared with baselines, validating practical impact for disaster-response and surveillance scenarios.

Abstract

UAV swarms are widely used in emergency communications, area monitoring, and disaster relief. Coordinated by control centers, they are ideal for federated learning (FL) frameworks. However, current UAV-assisted FL methods primarily focus on single tasks, overlooking the need for multi-task training. In disaster relief scenarios, UAVs perform tasks such as crowd detection, road feasibility analysis, and disaster assessment, which exhibit time-varying demands and potential correlations. In order to meet the time-varying requirements of tasks and complete multiple tasks efficiently under resource constraints, in this paper, we propose a UAV swarm based multi-task FL framework, where ground emergency vehicles (EVs) collaborate with UAVs to accomplish multiple tasks efficiently under constrained energy and bandwidth resources. Through theoretical analysis, we identify key factors affecting task performance and introduce a task attention mechanism to dynamically evaluate task importance, thereby achieving efficient resource allocation. Additionally, we propose a task affinity (TA) metric to capture the dynamic correlation among tasks, thereby promoting task knowledge sharing to accelerate training and improve the generalization ability of the model in different scenarios. To optimize resource allocation, we formulate a two-layer optimization problem to jointly optimize UAV transmission power, computation frequency, bandwidth allocation, and UAV-EV associations. For the inner problem, we derive closed-form solutions for transmission power, computation frequency, and bandwidth allocation and apply a block coordinate descent method for optimization. For the outer problem, a two-stage algorithm is designed to determine optimal UAV-EV associations. Furthermore, theoretical analysis reveals a trade-off between UAV energy consumption and multi-task performance.

Efficient UAV Swarm-Based Multi-Task Federated Learning with Dynamic Task Knowledge Sharing

TL;DR

The paper tackles multi-task federated learning for UAV swarms under strict energy and bandwidth limits, introducing a dynamic task attention mechanism and a task affinity metric to capture time-varying task importance and inter-task correlations. It develops a Lyapunov-based two-layer optimization to jointly optimize UAV transmission power, computation frequency, bandwidth allocation, and UAV–EV associations, with closed-form inner solutions and a two-stage outer association algorithm. The approach yields an O(\sqrt{V}, 1/V) trade-off between UAV energy consumption and multi-task performance and demonstrates improved convergence and generalization through targeted knowledge sharing across related tasks. Empirical results on MNIST, FMNIST, and FLAME-derived UAV datasets confirm faster convergence, higher accuracy, and better resource efficiency compared with baselines, validating practical impact for disaster-response and surveillance scenarios.

Abstract

UAV swarms are widely used in emergency communications, area monitoring, and disaster relief. Coordinated by control centers, they are ideal for federated learning (FL) frameworks. However, current UAV-assisted FL methods primarily focus on single tasks, overlooking the need for multi-task training. In disaster relief scenarios, UAVs perform tasks such as crowd detection, road feasibility analysis, and disaster assessment, which exhibit time-varying demands and potential correlations. In order to meet the time-varying requirements of tasks and complete multiple tasks efficiently under resource constraints, in this paper, we propose a UAV swarm based multi-task FL framework, where ground emergency vehicles (EVs) collaborate with UAVs to accomplish multiple tasks efficiently under constrained energy and bandwidth resources. Through theoretical analysis, we identify key factors affecting task performance and introduce a task attention mechanism to dynamically evaluate task importance, thereby achieving efficient resource allocation. Additionally, we propose a task affinity (TA) metric to capture the dynamic correlation among tasks, thereby promoting task knowledge sharing to accelerate training and improve the generalization ability of the model in different scenarios. To optimize resource allocation, we formulate a two-layer optimization problem to jointly optimize UAV transmission power, computation frequency, bandwidth allocation, and UAV-EV associations. For the inner problem, we derive closed-form solutions for transmission power, computation frequency, and bandwidth allocation and apply a block coordinate descent method for optimization. For the outer problem, a two-stage algorithm is designed to determine optimal UAV-EV associations. Furthermore, theoretical analysis reveals a trade-off between UAV energy consumption and multi-task performance.

Paper Structure

This paper contains 29 sections, 5 theorems, 84 equations, 13 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. Federated Learning System
  4. Computation Model
  5. Communication Model
  6. Problem Analysis And Problem Formulation
  7. Problem Analysis
  8. Convergence Analysis
  9. Proposed Algorithm
  10. Task Attention Mechanism
  11. Task Performance Balance
  12. Task Shapley Value
  13. Task Affinity
  14. Frequency and Power Control
  15. Optimize $\boldsymbol{f}$ given $\boldsymbol{p}$
  16. ...and 14 more sections

Key Result

Lemma 1

For any task $m,\forall m\in\mathcal{M}$, the difference between its loss functions in two consecutive rounds satisfies

Figures (13)

  • Figure 1: UAV swarm assisted multi-task FL scenario
  • Figure 2: The proposed scheme with task knowledge sharing, where EV1 shares knowledge with both EV2 and EV3, while EV3 shares knowledge with EV2.
  • Figure 3: RC-MNIST: 2Task
  • Figure 4: RC-MNIST: 3Task
  • Figure 6: SD-MNIST
  • ...and 8 more figures

Theorems & Definitions (5)

  • Lemma 1
  • Theorem 1
  • Lemma 2
  • Theorem 2
  • Theorem 3