Introduction
Unmanned Aerial Vehicles (UAVs) are increasingly deployed as "flying base stations" to provide on-demand wireless coverage in scenarios where terrestrial infrastructure is compromised, such as disaster zones, or overloaded, such as during massive public events. However, optimizing these networks involves navigating two major hurdles: the "backhaul bottleneck" and the diverse Quality-of-Service (QoS) requirements of ground users. This paper proposes a joint optimization framework that simultaneously tunes bandwidth allocation, transmit power, and flight trajectories to maximize network performance while adhering to these critical constraints.
System Model and the Backhaul Bottleneck
The network architecture consists of a UAV base station (BS) serving two distinct classes of ground users: delay-sensitive users (e.g., those on real-time video calls) and delay-tolerant users (e.g., those performing background file downloads). A central challenge in this model is the backhaul link, which connects the UAV-BS to the core network via a ground-based backhaul gateway.
Because spectrum is scarce, the backhaul link and the user data links share the same frequency bands. This creates a significant trade-off: if the UAV flies too far from the gateway to better serve a distant user, the backhaul link's capacity may drop so low that it cannot deliver the data the user requires, regardless of the strength of the UAV-to-user link.
Optimization Methodology
The authors aim to maximize the minimum average rate of delay-tolerant users, ensuring both fairness and high throughput. The resulting optimization problem is non-convex and computationally difficult due to the coupling of trajectory, power, and bandwidth variables. To find an efficient suboptimal solution, the paper employs two primary techniques:
- Alternating Optimization (AO): The problem is decomposed into two sub-blocks. The algorithm iteratively optimizes bandwidth and power for a fixed trajectory, and then updates the trajectory for the optimized resource allocation until the solution converges.
- Successive Convex Approximation (SCA): To handle the non-convexity of trajectory optimization—which involves complex distance-based constraints—SCA is used to approximate these constraints with simpler linear functions at each iteration, allowing for reliable pathfinding.
Key Insights and Discussion
- The "Tether" Effect: The UAV's optimal flight path is visibly "pulled" toward the backhaul gateway. It must find a compromise position that maintains a strong link to the gateway while staying close enough to the users to provide high-quality service.
- Successive Service (Spatial TDMA): The optimization reveals that the most efficient way to serve delay-tolerant users is often one at a time. The UAV tends to fly toward one user, provide them with the majority of resources, and then move on to the next—effectively a spatial version of Time Division Multiple Access (TDMA).
- QoS Anchoring: Delay-sensitive users act as spatial anchors. Because they require a guaranteed minimum bit rate at every moment, the UAV is restricted from flying too far away from them, even if doing so would allow it to reach a delay-tolerant user more efficiently.
Simulation Results
The framework was validated using computer-aided simulations in a $2 \times 2 \text{ km}^2$ area with six ground users (two delay-sensitive and four delay-tolerant). The UAV operated at a fixed altitude of 100m with a maximum speed of 50 m/s using a standard Line-of-Sight (LoS) channel model.
Key findings include:
- Superior Performance: The proposed joint optimization significantly outperformed benchmark schemes, such as those using fixed circular trajectories or static bandwidth allocations.
- Resource Synergy: While optimizing power alone provides only marginal gains, the real performance "heavy lifting" comes from the joint optimization of trajectory and bandwidth.
- Flight Time Impact: As the total flight duration ($T$) increases, the UAV has more freedom to visit each user. Data rates improve with $T$ but eventually hit a plateau dictated by the fundamental capacity of the backhaul link.
Limitations and Future Work
While the current model provides a robust framework, it relies on a simplified Line-of-Sight (LoS) channel model. In dense urban environments with skyscrapers, signal fluctuations would require more complex probabilistic models. Furthermore, the current study assumes a fixed altitude and a single UAV. Future research will explore 3D trajectory optimization and multi-UAV coordination to avoid interference while sharing common backhaul resources.
Summary updated