Joint D2D Assignment, Bandwidth and Power Allocation in Cognitive UAV-Enabled Networks

Huy T. Nguyen, Hoang Duong Tuan, Trung Q. Duong, H. Vincent Poor, Won-Joo HwangIEEE Trans. Cogn. Commun. Netw 2020 · DOI

Abstract

As wireless networks evolve toward 5G and beyond, Unmanned Aerial Vehicles (UAVs) are increasingly utilized as "flying base stations" to enhance coverage in rural or emergency scenarios. However, integrating these with Device-to-Device (D2D) communications introduces significant interference management challenges. This paper explores a joint optimization framework for D2D assignment, bandwidth allocation, and power control in cognitive UAV-enabled networks. By transforming a non-convex mixed-binary problem into a continuous one using an exact penalty reformulation, the authors develop an efficient alternating descent algorithm. Simulation results indicate that this approach maximizes system throughput while maintaining interference constraints, revealing that lower UAV altitudes can surprisingly benefit network performance.

Introduction

The deployment of UAVs as aerial base stations offers a flexible solution for providing high-speed wireless access. In these networks, D2D users often coexist with the UAV's ground terminals (GTs) by reusing the same frequency bands to improve spectral efficiency. This "cognitive" setup, however, creates a complex interference environment. The strong Line-of-Sight (LoS) links from the UAV can disrupt D2D communications, while D2D transmissions can conversely interfere with the UAV-to-GT links.

System model of a cognitive UAV-enabled network showing communication and interference links.

To optimize such a network, three interdependent decisions must be made simultaneously: which GT's spectrum a D2D pair should share (D2D assignment), how to divide the total bandwidth among GTs, and how to allocate transmit power for both the UAV and D2D users. Combining these creates a mixed-binary optimization problem that is notoriously difficult (NP-hard) to solve in real-time.

Methodology

The authors propose a sophisticated mathematical framework to solve this joint design problem efficiently:

  • Exact Penalty Reformulation: To handle the discrete (0 or 1) nature of D2D assignment, a novel penalty function is introduced. This allows the binary variables to be treated as continuous during the optimization process, with a penalty applied if they do not converge to binary values, smoothing the mathematical landscape.
  • Alternating Descent Algorithm: The joint problem is decoupled into two manageable sub-problems: one for D2D assignment and another for joint bandwidth and power allocation. These are solved iteratively to reach a global solution.
  • Convex Approximation: Non-convex logarithmic terms in the objective function are handled using "difference of convex" (d.c.) programming, approximating them into solvable convex problems at each iteration.

The framework was validated through numerical simulations in a virtual cell with a 500m radius, featuring 5 GTs and 20 D2D pairs. The researchers utilized standard LoS and Rayleigh fading channel models to test the theoretical limits of the algorithm.

Key Insights and Results

The study yielded several critical findings regarding the behavior of cognitive UAV networks:

  • The "Lowest is Best" Paradox: Contrary to the intuition that higher altitude provides better coverage, the results showed that the network performs best when the UAV hovers at its lowest possible altitude. This configuration strengthens the intended LoS signal more significantly than it increases interference to D2D users, while remaining more energy-efficient.
  • Efficiency of Spectrum Sharing: The cognitive approach (sharing bandwidth) consistently outperformed non-cognitive methods (dedicated bandwidth slices), provided that the proposed interference management algorithm was employed to balance the needs of all users.
  • Algorithm Performance: The proposed algorithm demonstrated rapid convergence, typically reaching an optimal solution within 6 to 10 iterations. Furthermore, the power control mechanism successfully protected D2D users by saturating the UAV's transmit power (around 0.6W) even when more power was available.

Conclusion

This research provides a robust framework for managing the complexities of UAV-D2D coexistence. By effectively balancing bandwidth and power across the network, the proposed optimization significantly boosts throughput. Future work will look to extend this model by incorporating UAV trajectory optimization and addressing the challenges of imperfect channel state information (CSI) in dynamic real-world deployments.

Summary updated

BibTeX
@article{nguyen2020joint,
  author = {Huy T. Nguyen and Hoang Duong Tuan and Trung Q. Duong and H. Vincent Poor and Won-Joo Hwang},
  title = {{Joint D2D Assignment, Bandwidth and Power Allocation in Cognitive UAV-Enabled Networks}},
  journal = {IEEE Trans. Cogn. Commun. Netw.},
  year = {2020},
  doi = {10.1109/tccn.2020.2969623},
}
Figure