Joint Coverage and Power Control in Highly Dynamic and Massive UAV\n Networks: An Aggregative Game-theoretic Learning Approach

Jing Wu, Qimei Chen, Hao Jiang, Haozhao WangarXiv 2019 · DOI · arXiv

Abstract

Unmanned aerial vehicle (UAV) ad-hoc networks serve as critical contingency plans for communication after natural disasters. To achieve efficient deployment, this paper employs noncooperative game theory and a novel synchronous payoff-based binary log-linear learning algorithm (SPBLLA). By modeling the network as an aggregative game, the approach accounts for channel overlap, power control, and coverage complexity while addressing the limitations of large-scale deployments in dynamic environments. Experimental results indicate that SPBLLA significantly reduces information exchange and achieves a learning rate approximately 10 times faster than traditional asynchronous methods, making it highly suitable for urgent disaster response scenarios.

Introduction

When natural disasters destroy ground-based cell towers, UAVs can be deployed as an "ad-hoc" network to restore connectivity. However, managing a massive network of hundreds of drones in a highly dynamic environment—characterized by wind, rain, and moving users—is a significant challenge for traditional control systems.

The topological structure and communication links in a post-disaster UAV network scenario.

The authors identify three main gaps in current research: the computational complexity of tracking individual drone interactions in large swarms, the high coordination overhead of asynchronous algorithms, and the information constraints where drones cannot easily sense the exact states of their neighbors.

Methodology

Aggregative Game Theory

Instead of modeling the network as a complex web of individual interactions, the researchers use an aggregative game. In this model, a drone does not need to track the specific strategy of every other drone. Instead, it focuses on the aggregate interference and total coverage area of the swarm. This reduces the data processing requirement from tracking $N$ individual strategies to monitoring just a few aggregate values, allowing for more efficient onboard processing.

Illustration of users supported by different UAVs and the resulting coverage overlap.

SPBLLA: Synchronous Payoff-Based Learning

The core innovation is the Synchronous Payoff-Based Binary Log-Linear Learning Algorithm (SPBLLA). Unlike traditional algorithms that require drones to update one at a time to maintain stability, SPBLLA allows multiple drones to update their power and altitude levels simultaneously. It is "payoff-based," meaning drones optimize their behavior by comparing their current performance (a mix of signal quality and energy efficiency) with previous states, rather than requiring constant communication with neighbors.

Experimental Evaluation

The study utilized numerical simulations of a $4000 \text{ km}^2$ post-disaster area with up to 200 identical UAVs. The network configuration included 30–50 communication channels and dozens of power and altitude levels.

Key results include:

  • Speedup: SPBLLA converged to an optimal state roughly 10 times faster than the traditional asynchronous BLLA.
  • Coverage: The swarm successfully covered 95% of the disaster area in low-turbulence scenarios and maintained 80% coverage even in highly dynamic, noisy environments.
  • Scalability: Global utility increased linearly with the number of UAVs, proving the model's effectiveness for massive deployments.

Conclusion

By combining aggregative game theory with synchronous learning, this research provides a promising model for large-scale, highly dynamic UAV networks. The SPBLLA algorithm effectively balances the need for rapid deployment with the physical constraints of drone hardware. Future work will look into joint optimization where drones can switch channels, altitudes, and power levels simultaneously to further enhance network resilience.

Summary updated

Abstract

Unmanned aerial vehicles (UAV) ad-hoc network is a significant contingency\nplan for communication after a natural disaster, such as typhoon and\nearthquake. To achieve efficient and rapid networks deployment, we employ\nnoncooperative game theory and amended binary log-linear algorithm (BLLA)\nseeking for the Nash equilibrium which achieves the optimal network\nperformance. We not only take channel overlap and power control into account\nbut also consider coverage and the complexity of interference. However,\nextensive UAV game theoretical models show limitations in post-disaster\nscenarios which require large-scale UAV network deployments. Besides, the\nhighly dynamic post-disaster scenarios cause strategies updating constraint and\nstrategy-deciding error on UAV ad-hoc networks. To handle these problems, we\nemploy aggregative game which could capture and cover those characteristics.\nMoreover, we propose a novel synchronous payoff-based binary log-linear\nlearning algorithm (SPBLLA) to lessen information exchange and reduce time\nconsumption. Ultimately, the experiments indicate that, under the same\nstrategy-deciding error rate, SPBLLA's learning rate is manifestly faster than\nthat of the revised BLLA. Hence, the new model and algorithm are more suitable\nand promising for large-scale highly dynamic scenarios.\n
BibTeX
@misc{wu2019joint,
  author = {Jing Wu and Qimei Chen and Hao Jiang and Haozhao Wang},
  title = {{Joint Coverage and Power Control in Highly Dynamic and Massive UAV\textbackslash\{\}n Networks: An Aggregative Game-theoretic Learning Approach}},
  year = {2019},
  eprint = {1907.08363},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1907.08363},
}
Figure