Source-linked AI summary
Backhaul-aware Robust 3D Drone Placement in 5G+ Wireless Networks
Elham Kalantari, Muhammad Zeeshan Shakir, Halim Yanikomeroglu, Abbas Yongacoglu
TL;DR
Drone-BS placement must account for limited and weather-sensitive wireless backhaul while serving users with heterogeneous rate requirements. The paper develops backhaul-aware 3D placement under network-centric and user-centric objectives, finding robustness to modest user movement and a backhaul-dependent service limit.
Problem
Wireless backhaul can constrain drone-BS peak data rates and may degrade under adverse weather, motivating its inclusion as a deployment-design constraint.
Method
The paper proposes a backhaul-limited optimal 3D placement algorithm that maximizes served-user count or sum-rate under heterogeneous rate requirements, and evaluates robustness to user movement.
Results
Less than 2% of users moving within 100 meters would be disconnected in the network-centric approach, versus less than 1% in the user-centric approach.
Takeaways & Limitations
Network-centric placement serves more users than user-centric placement, while increasing backhaul capacity raises served users until about 150 Mbps.
Takeaways & Limitations
The backhaul link may be dedicated or in-band, so the deployment setting includes distinct backhaul configurations.
Abstract
from arXiv · showhide
Using drones as flying base stations is a promising approach to enhance the network coverage and area capacity by moving supply towards demand when required. However deployment of such base stations can face some restrictions that need to be considered. One of the limitations in drone base stations (drone-BSs) deployment is the availability of reliable wireless backhaul link. This paper investigates how different types of wireless backhaul offering various data rates would affect the number of served users. Two approaches, namely, network-centric and user-centric, are introduced and the optimal 3D backhaul-aware placement of a drone-BS is found for each approach. To this end, the total number of served users and sum-rates are maximized in the network-centric and user-centric frameworks, respectively. Moreover, as it is preferred to decrease drone-BS movements to save more on battery and increase flight time and to reduce the channel variations, the robustness of the network is examined as how sensitive it is with respect to the users displacements.
I. INTRODUCTION
Drone-BSs provide temporary capacity or coverage, but their deployment must account for constrained wireless backhaul. The paper proposes backhaul-limited placement for different objectives and evaluates robustness to user movement.
- I. INTRODUCTION: Drone-BSs can temporarily increase capacity during congestion and provide coverage when terrestrial infrastructure is damaged or unavailable.Use cases include temporary events, remote or sparsely populated areas, and outages caused by weather, vandalism, or transmission problems.
- I. INTRODUCTION: Wireless backhaul is a central drone-BS constraint because available peak data rates can be limited and may decrease under adverse weather.The limitation is especially relevant for wireless links using FSO or mmWave technologies.
- I. INTRODUCTION: The paper proposes a backhaul-limited optimal 3D drone-BS placement algorithm for maximizing served users or sum-rate with heterogeneous rate requirements.The design considers clustered user distributions and different network parameters.
- I. INTRODUCTION: The study evaluates how user movement affects the proposed optimal placement and examines the robustness of the resulting solution.This complements the placement design by testing sensitivity to user displacements.
II. SYSTEM MODEL
The system model represents air-to-ground propagation through probabilistic line-of-sight and non-line-of-sight connections. Average pathloss combines free-space loss with environment-dependent additional losses, without modeling shadowing.
- A. Pathloss Model: The air-to-ground model distinguishes line-of-sight and non-line-of-sight propagation and assigns a probability to line-of-sight connections.The adopted model accounts for signals received through strong reflections and diffractions in non-line-of-sight conditions.
- A. Pathloss Model: The line-of-sight probability depends on environment-specific constants and the elevation angle, defined from drone altitude and horizontal receiver distance.The elevation angle is θ = arctan(h/r), where h is altitude and r is horizontal distance.
- A. Pathloss Model: Average pathloss is modeled probabilistically without shadowing, using free-space pathloss together with additional line-of-sight and non-line-of-sight losses.The additional losses depend on the respective environment.
B. Spatial Users Distribution
The paper models heterogeneous user locations with a Matérn cluster process, placing users uniformly within circular clusters around Poisson-distributed parent points.
- A Matérn cluster process generates heterogeneous user distributions through parent cluster centers and daughter users.Parent points follow a homogeneous Poisson process, while users are uniformly scattered within circles of radius ν around the parents using another Poisson process.
III. BACKHAUL-AWARE DRONE-BS PLACEMENT
The proposed placement framework integrates a drone-BS with an existing terrestrial network to serve otherwise unsupported users under backhaul, bandwidth, coverage, and placement constraints.
- The drone-BS supplements ground-BSs when temporary demand or rate increases exceed terrestrial network resources.Users have heterogeneous application rate requirements, and the selected users depend on whether the framework is network-centric or user-centric.
- The wireless backhaul constrains the aggregate served-user rate to the link capacity R Mbps.NU is the number of users not served terrestrially, ri is user i’s required rate, and Ii indicates whether the drone-BS serves that user.
- The drone-BS also faces a total bandwidth constraint determined by users’ required bandwidths.Each user’s bandwidth requirement depends on spectral efficiency ζi = log2(1 + γi), where γi is the user’s SNR.
- Users are considered covered when their received pathloss does not exceed the maximum tolerable pathloss imposed by QoS.PLi denotes the received pathloss for user i, while PLmax is the QoS-dependent outage threshold.
- The centralized algorithm exhaustively searches candidate 3D coordinates and solves a binary integer linear program at each placement.The placement coordinates are bounded in x, y, and h, with altitude limits influenced by drone characteristics and regulations; simulations use a 16 km2 urban region and 100-meter search steps.
A. Network-Centric versus User-Centric
The framework selects users using either a network-centric objective that prioritizes user count or a user-centric objective that incorporates user priorities.
- The network-centric approach maximizes the number of served users regardless of their rate requirements.It sets αi = 1 for every user, so users requiring lower data rates tend to comprise most of the served set.
1) Sum-Rate:
The user-centric approach uses total sum-rate as its selection metric, prioritizing users with higher required data rates.
- Setting αi equal to ri gives higher-rate users higher priority in network access.The paper uses total sum-rate maximization as its user-centric metric.
2) Price Differentiation:
Users may be prioritized by subscription value, with higher-paying users receiving stronger service priority. The figure presents user distribution and drone-BS placement for the two approaches.
- 2) Price Differentiation:: The figure shows user distribution and 3D drone-BS placement for network-centric and user-centric approaches.The drone-BS is marked by an asterisk, while its XY-plane projection is shown by red circles.
- 2) Price Differentiation:: Platinum users receive priority because they expect connectivity under nearly every condition.The model represents this priority by assigning them a large α_i value.
3) Signal Strength:
User selection can prioritize favorable signal conditions or urgent content needs. The network-centric placement serves more users than the user-centric placement, partly because it serves users requiring lower rates more often.
- 3) Signal Strength:: User selection may prioritize users with favorable received signal strength.This approach serves users with better channel conditions first.
- 3) Signal Strength:: Content-aware selection gives higher priority to users who urgently need particular content.
- 3) Signal Strength:: The network-centric placement serves more users than the user-centric placement while both place the drone-BS at hmax.The higher altitude expands the covered area; the network-centric approach can also reduce spectrum-usage fees by serving more users per area.
- 3) Signal Strength:: The network-centric CDF lies above the user-centric CDF, indicating a higher probability of serving users with lower required rates.This rate distribution is consistent with the larger total number of served users in the network-centric approach.
B. Backhaul Limitation
Wireless backhaul can be dedicated or in-band, with different capacity and reliability characteristics. Low backhaul rates constrain service, while increasing capacity raises served users until drone-BS spectrum becomes limiting near 150 Mbps.
- B. Backhaul Limitation: Dedicated FSO or mmWave backhaul offers very high capacity but is highly sensitive to weather conditions.Fog or rain can dramatically reduce the peak rate.
- B. Backhaul Limitation: RF microwave backhaul deploys quickly at relatively low cost but reuses access-link spectrum and suffers interference.The shared spectrum also affects backhaul connection capacity.
- B. Backhaul Limitation: Low wireless backhaul rates can severely limit the number of served users.The compared rates represent different types of wireless links.
- B. Backhaul Limitation: Served users stop increasing when backhaul capacity reaches around 150 Mbps because drone-BS spectrum resources are exhausted.
- B. Backhaul Limitation: User-centric service has an almost fixed increase rate, whereas network-centric service gains diminish as backhaul capacity increases.High-rate-first selection explains the fixed user-centric slope; low-rate-first selection leaves only a few high-rate users to add in the network-centric case.
C. Robustness
Keeping the drone-BS stationary reduces movement-related energy use and channel-management complexity. The proposed placement remains robust when users move modest distances, with only a small reduction in service.
- C. Robustness: Constant drone-BS movement consumes battery and reduces flight time while continually changing the radio channel.A predetermined position can reduce interference-management and resource-allocation complexity.
- C. Robustness: Increasing user movement distance decreases served users, but the reduction is not significant.The corresponding figure evaluates remaining served users and the percentage of dropped-out users.
- C. Robustness: 100-meter user movement disconnects less than 2% of users in the network-centric approach and less than 1% in the user-centric approach.The authors therefore characterize the placement as robust for this movement range.
- C. Robustness: The drone-BS can remain in a suitable position until a predetermined user-dropout threshold is reached.
- C. Robustness: The study reports that only a small percentage of served users enter outage when users move within a few meters.This supports robustness against modest user movement.