Abstract
The use of drone base stations to provide wireless connectivity for ground terminals is becoming a promising part of future technologies. The design of such aerial networks is however different compared to cellular 2D networks, as antennas from the drones are looking down, and the channel model becomes height-dependent. In this paper, we study the effect of antenna patterns and height-dependent shadowing. We consider a random network topology to capture the effect of dynamic changes of the flying base stations. First we characterize the aggregate interference imposed by the co-channel neighboring drones. Then we derive the link coverage probability between a ground user and its associated drone base station. The result is used to obtain the optimum system parameters in terms of drones antenna beamwidth, density and altitude. We also derive the average LoS probability of the associated drone and show that it is a good approximation and simplification of the coverage probability in low altitudes up to 500 m according to the required signal-to-interference-plus-noise ratio (SINR).
Introduction
Deploying drones as temporary base stations is a promising solution for disaster recovery or high-traffic events. However, designing these aerial networks is significantly more complex than traditional 2D ground-based cellular networks. Drones operate in 3D space where the radio channel is highly dependent on altitude. As a drone flies higher, the probability of a clear Line-of-Sight (LoS) to a user increases, but the signal strength decreases due to greater distance. Furthermore, in a multi-drone network, users face interference from all neighboring drones. This research addresses the dynamic, "on-the-fly" nature of drone deployments by modeling them as random fields.
System Model and Methodology
The authors utilized Stochastic Geometry to model the inherent randomness of drone locations, treating them as a Poisson Point Process (PPP). This mathematical approach allows for the analysis of networks where drone positions are not fixed or known in advance.
Key technical components of the model include:
- Height-Dependent Channel Model: Accounting for changes in LoS probability and shadowing based on elevation.
- Directional Antenna Modeling: Drones are modeled with specific downward-pointing antenna beamwidths.
- Closest-Drone Association: Ground users are assumed to connect to the nearest available drone.
- Analytical Derivations: The study derives closed-form expressions for mean aggregate interference and coverage probability—the likelihood that the signal-to-interference-plus-noise ratio (SINR) exceeds a required threshold.
Key Results and Optimization
The analysis reveals several critical insights for network planning:
The Altitude "Sweet Spot"
There is a distinct optimal altitude for drone deployment. At low altitudes, signal blockage by buildings (low LoS) limits performance. At high altitudes, path loss due to distance becomes the dominant factor.
Interference Invariance and LoS Approximation
A significant finding is that average interference from neighboring drones remains relatively constant across altitudes. While higher drones have more interfering neighbors in view, the signal from those neighbors is weaker. Additionally, the researchers found that for altitudes below 500m, the complex coverage probability can be accurately approximated by the average LoS probability of the closest drone, simplifying network optimization.
Beamwidth and Density Trade-offs
The study identifies optimal antenna beamwidths (typically 60°–100°) and drone densities. While increasing drone density initially improves coverage by bringing users closer to base stations, excessive density leads to a "crowded" spectrum where interference causes coverage to drop sharply.
Conclusion
This paper provides a mathematical blueprint for optimizing drone-based cellular networks. By identifying the optimal combinations of altitude, density, and beamwidth, operators can maximize coverage in various environments, from suburban to high-rise urban areas. Future work will extend this model to include hybrid networks with terrestrial towers, non-uniform drone arrangements, and energy efficiency considerations.
Summary updated