Source-linked AI summary
On the Number and 3D Placement of Drone Base Stations in Wireless Cellular Networks
Elham Kalantari, Halim Yanikomeroglu, Abbas Yongacoglu
TL;DR
The paper asks how to minimize the number of drone-BSs while placing them in 3D to serve users with changing densities and QoS requirements. It formulates coverage and capacity constraints and uses a PSO-based heuristic with redundant-station removal. Simulations show acceptable performance across regions with different user densities, including density-dependent counts and altitude adjustments.
Problem
The paper addresses how to find the minimum number and placement of drone-BSs needed to provide target QoS coverage to arbitrarily located users.
Method
A modified particle swarm optimization heuristic estimates drone-BS number and 3D locations under coverage and capacity constraints, then removes redundant stations.
Results
Simulations across regions with different user densities confirmed acceptable performance, with drone-BS counts proportional to local user density and altitudes adjusted for coverage and capacity.
Takeaways & Limitations
Drone-BSs can be deployed with density-dependent placement and altitude changes while satisfying the system’s coverage and capacity constraints.
Abstract
from arXiv · showhide
Using drone base stations (drone-BSs) in wireless networks has started attracting attention. Drone-BSs can assist the ground BSs in both capacity and coverage enhancement. One of the important problems about integrating drone-BSs to cellular networks is the management of their placement to satisfy the dynamic system requirements. In this paper, we propose a method to find the positions of drone-BSs in an area with different user densities using a heuristic algorithm. The goal is to find the minimum number of drone-BSs and their 3D placement so that all the users are served. Our simulation results show that the proposed approach can satisfy the quality-of-service requirements of the network.
I. INTRODUCTION
Drone-BSs are motivated by unpredictable spatial and temporal demand, offering a potentially more flexible alternative to dense ground-BS deployment. The paper addresses the open problem of minimizing their number while jointly determining 3D placements that meet user QoS.
- Motivation: Dense ground-BS deployment is costly because many stations may be lightly loaded or unloaded as user demand varies across space and time.Traffic patterns are difficult to predict because user densities and application rates vary temporally and spatially.
- Motivation: Drone-BSs can provide high-data-rate coverage when excessive demand occurs in difficult-to-predict locations and times.
- Related work: Prior work studies air-to-ground pathloss, line-of-sight probability, UAV-assisted D2D performance, and optimal altitude or coverage for one or two drone-BSs.
- Contribution: The paper targets the previously unaddressed problem of finding the minimum number of drone-BSs and their placements to cover users with target QoS.
II. SYSTEM MODEL
The system model considers one snapshot of user positions and seeks the minimum number and 3D placement of downlink drone-BSs needed to serve those users.
- The objective is to find the minimum number of transmitting drone-BSs and their 3D placement for users in a specified area.
- Because drone-BSs can move, the model need not cover areas without users and represents placement for one snapshot of user positions.The stations might follow users as they move.
A. Air-to-ground channel model
The air-to-ground model uses line-of-sight probability and mean pathloss, with altitude affecting both propagation conditions and transmitter–receiver distance. Consequently, pathloss can decrease and then increase as altitude rises.
- The air-to-ground channel differs from terrestrial models, and line-of-sight probability is an important pathloss factor.
- For fixed horizontal distance, increasing altitude raises the elevation angle and increases the probability of a line-of-sight connection.The elevation angle is defined using drone-BS altitude and horizontal receiver distance.
- The adopted model excludes random radio-channel behavior and uses a mean pathloss model.
- Changing drone-BS height can attain a desired coverage area without changing transmit power.
B. Optimization problem
The optimization problem combines coverage and capacity requirements to place drone-BSs for users distributed across equal-density subareas. The formulation starts from an estimated station count and supports iterative removal of redundant stations.
- B. Optimization problem: The initial drone-BS count is estimated from coverage and capacity requirements, with user traffic mainly driving the need for multiple stations.The coverage model allows one optimally elevated drone-BS to cover a very large area.
- B. Optimization problem: Each drone-BS has capacity CBS = B×η, where B is total bandwidth and η is average system spectral efficiency.The target download rate is R, and the floor function is used in the station-count estimate.
- B. Optimization problem: The coverage constraint requires at least ζ percent of users to be covered by drone-BSs.
- B. Optimization problem: The capacity constraint allocates each drone-BS’s capacity across subareas according to mutual-area fractions ρj,k = aj,k/Aj.A subarea is a region in which user density is the same, and ρj,k lies between zero and one.
- B. Optimization problem: The formulation includes constraints for subarea demand, average spectral efficiency, three-dimensional drone-BS positions, and indicators for used or redundant stations.
III. PROPOSED ALGORITHM
The paper formulates drone-BS placement as a complex 3D optimization problem and uses a modified PSO algorithm to find feasible locations. It then minimizes the number of deployed drone-BSs while preserving service constraints.
- 3D placement is complicated because the optimization includes drone-BS altitude as an additional dimension.
- The algorithm sequentially optimizes capacity, coverage, and additional utility constraints to produce a feasible placement.The placement is improved by minimizing uncovered users while retaining the capacity constraint.
- PSO searches for drone-BS locations using random particles, personal bests, and the global best solution.Each particle encodes the positions of all drone-BSs.
- Each particle’s velocity is updated from inertia and personal and global learning terms before its position is updated.The inertia weight controls convergence speed, while the learning coefficients weight personal and global information.
- After obtaining a feasible placement serving ζ percent of users by traffic requirements, the algorithm removes dispensable drone-BSs iteratively.A drone-BS is removed only when the network constraints continue to hold.
IV. SIMULATION RESULTS
Simulations evaluate PSO-based drone-BS placement under two user-density patterns in a 100 km2 urban area. The resulting placements adapt drone-BS number, spacing, and altitude to regional density while reporting SINR distributions and algorithm convergence.
- 100 km2 and 1000 users define the simulated urban deployment area.
- Scenario I: Scenario I divides users uniformly between two equal regions containing 20% and 80% of users.The initial estimate uses 30 drone-BSs, with one ultimately identified as redundant.
- SINR distributions and PSO convergence are reported for both scenarios.Figures 4–5 cover Scenario I, while Figures 7–8 cover Scenario II.
- Scenario I: Higher-density regions receive more closely placed drone-BSs and lower altitudes to serve nearby users and reduce interference.In Scenario I, lower density also permits higher altitudes for broader coverage and reduced pathloss.
- Scenario II: Scenario II assigns 40% of users to a central normal distribution and 60% uniformly to the remaining region.The final placement requires 29 drone-BSs to serve all users.
V. CONCLUSION
The paper addresses drone-BS deployment as a complex optimization problem requiring both the minimum number of stations and their 3D locations. It proposes PSO-based heuristic placement with coverage and capacity constraints, then removes redundant stations; simulations support acceptable performance across different user densities.
- The study targets minimum drone-BS count and 3D placement for serving arbitrarily located users with QoS.
- Because the optimization is too complex for direct solution, the paper proposes a heuristic algorithm based on particle swarm optimization.
- The method estimates station count and 3D placement under coverage and capacity constraints, then removes stations whose removal does not affect network quality.
- Simulations across different user densities report acceptable performance, with station count proportional to regional user density.
- Drone-BSs lower altitude in dense areas to reduce interference and raise altitude in sparse areas to cover larger regions.