⚡ electrical engineering

Joint Optimization of Multi-UAV Deployment and 3D Positioning in Traffic-Aware Aerial Networks

This paper proposes the Efficient Multi-UAV Traffic-Aware Deployment (EMTAD) algorithm, a scalable framework that jointly optimizes the 3D positioning and number of UAVs in real-time to meet dynamic traffic demands while maximizing network throughput and minimizing deployment overhead.

Original authors: Kamran Shafafi, Alaa Awad Abdellatif, Manuel Ricardo, Rui Campos

Published 2026-01-30

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Kamran Shafafi, Alaa Awad Abdellatif, Manuel Ricardo, Rui Campos

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a city where the ground is crowded with people trying to send messages, videos, and data. Sometimes, the buildings block the signal, making it hard to get a clear connection. To fix this, we send up a fleet of flying drones (UAVs) to act as temporary cell towers in the sky.

The problem is: How do you position these drones?

If you just drop them randomly or keep them in one spot, you might waste money using too many drones, or worse, leave some people with no signal because the traffic is too heavy for one drone to handle.

This paper introduces a smart system called EMTAD (Efficient Multi-UAV Traffic-Aware Deployment). Think of it as a highly intelligent traffic controller for the sky. Here is how it works, broken down into simple concepts:

1. The "Bubble" Concept

Every person on the ground has a specific amount of data they need to send (like streaming a movie vs. sending a text).

The Analogy: Imagine every person is standing inside an invisible, floating bubble. The size of this bubble depends on how much data they need. If they need a lot of data, the bubble is small (the drone must be close). If they need little data, the bubble is large (the drone can be far away).
The Goal: The system tries to find a spot in the sky where one drone can sit inside the bubbles of multiple people at the same time.

2. The "Intersection" Strategy

Instead of guessing where to put the drones, the EMTAD algorithm looks for the "sweet spots" where these invisible bubbles overlap.

The Analogy: Imagine you have several people, each with their own bubble. You want to find a single spot where you can place a drone that is close enough to serve Person A, Person B, and Person C all at once. The algorithm finds these overlapping zones (intersections).
The Result: By placing a drone in these overlapping zones, one drone can serve many people, meaning you need fewer total drones to cover the whole crowd.

3. The "Smart Swarm" (PSO)

Once the algorithm finds the best zones, it needs to figure out the exact 3D position (height and location) for each drone to get the best signal.

The Analogy: Think of a swarm of birds looking for the best place to land. They don't just guess; they test different spots, see which one gives the best connection, and move slightly to improve it. This is called Particle Swarm Optimization (PSO). The algorithm uses this "swarm" logic to fine-tune the drone's position until it finds the perfect spot to maximize speed and minimize interference.

4. What the Paper Found

The researchers tested this system in three different "simulated cities" using computer models:

Scenario A (Different Traffic): When people needed different amounts of data (some streaming video, some just texting), EMTAD used fewer drones than older methods while still keeping everyone happy.
Scenario B (Different City Sizes): Whether the area was small or huge, EMTAD adjusted the number of drones needed. In smaller areas, it often needed just one drone, whereas older methods tried to use two.
Scenario C (More People): As the crowd grew, EMTAD scaled up efficiently. It managed to serve a crowd of 60 people with only two drones, while the old methods needed up to six drones to do the same job.

The Bottom Line

The main achievement of this paper is that EMTAD saves resources. It doesn't just put drones in the sky; it calculates the minimum number of drones needed to do the job perfectly. It ensures that:

No one is left behind: Everyone gets their required data speed.
No drones are wasted: It doesn't deploy a drone if one isn't strictly necessary.
The signal is clear: It positions drones to avoid buildings and maximize "Line-of-Sight" (a direct, unblocked path for the signal).

In short, this paper presents a smarter way to fly drones so that we can get better internet in busy cities without needing a massive, expensive fleet of aircraft.

Technical Summary: Joint Optimization of Multi-UAV Deployment and 3D Positioning in Traffic-Aware Aerial Networks

Problem Statement
The paper addresses the critical challenge of deploying multiple Unmanned Aerial Vehicles (UAVs) in dynamic, traffic-aware aerial networks. While UAVs offer on-demand deployment and reliable Line-of-Sight (LoS) connectivity, effectively positioning them in real-time to meet non-uniform, time-varying traffic demands remains difficult. Existing methods often rely on static user distributions, fixed numbers of UAVs, or separate optimization stages that fail to jointly minimize the number of deployed UAVs while maximizing network throughput. The core problem is formulated as a Mixed-Integer Nonlinear Programming (MINLP) problem: simultaneously determining the minimum number of active UAVs, their optimal 3D coordinates, and the association of ground User Equipment (UEs) to these UAVs, subject to data rate requirements, bandwidth constraints, and LoS probability models. The authors note that this problem is NP-hard, rendering optimal solutions computationally intractable for large-scale scenarios.

Methodology: The EMTAD Algorithm
To solve this optimization problem, the authors propose the Efficient Multi-UAV Traffic-Aware Deployment (EMTAD) algorithm. This framework operates in two primary stages:

Spatial Clustering via Coverage Intersection:
- The algorithm first calculates a maximum allowable distance ( $d^{MAX}_{ik}$ ) between a UE and a UAV. This distance is derived from the user's specific traffic demand, the allocated bandwidth, and the probabilistic LoS/NLoS channel model (based on the Al-Hourani model for urban environments).
- Based on this distance, a "coverage sphere" is defined around each UE.
- The algorithm computes the spatial intersections of these spheres to identify candidate 3D regions ( $s_p$ ) where a single UAV can simultaneously serve multiple UEs while satisfying their individual traffic constraints.
- It then identifies the smallest subset of these intersection zones required to cover all UEs, effectively grouping users based on proximity and traffic compatibility to minimize the total number of UAVs needed.
3D Positioning Optimization via Particle Swarm Optimization (PSO):
- Once the candidate intersection zones are established, a PSO-based metaheuristic is employed to determine the precise 3D coordinates ( $P_k$ ) for each UAV within its assigned zone.
- The PSO fitness function aims to maximize the number of UEs served per UAV (adhering to bandwidth limits) while ensuring the achievable data rate meets the user's demand and prioritizing positions with high LoS probability.
- The algorithm iteratively reduces the number of UAVs ( $K_{max}$ ) to find the most efficient deployment configuration.

Key Contributions
The paper outlines three main contributions:

Traffic-Aware Real-Time Positioning: A strategy that dynamically adjusts UAV placements based on real-time user locations and spatial traffic distribution, rather than static patterns.
Integrated Optimization Framework: A unified approach that jointly minimizes the number of deployed UAVs, determines optimal UE-UAV associations, and accounts for practical constraints such as bandwidth capacity and LoS propagation.
Scalable Algorithm: A heuristic solution designed for dynamic, heterogeneous network environments that balances resource efficiency with overall network performance.

Performance Evaluation and Results
The authors evaluated EMTAD using Python scripts and the ns-3 network simulator, comparing it against two state-of-the-art baselines: a fixed-altitude approach and a fixed-number-of-UEs-per-UAV approach. The evaluation covered three scenarios:

Scenario A (Varying Traffic): With a fixed number of UEs (20) and venue size (100m x 100m), EMTAD demonstrated superior performance in aggregate throughput compared to baselines, particularly when higher traffic demands required more UAVs.
Scenario B (Varying Venue Size): In larger venues (up to 500m x 500m), EMTAD achieved comparable or better throughput with fewer UAVs. For instance, in a 500m x 500m area, EMTAD deployed three UAVs to meet demands, whereas the benchmark approach with two UAVs failed to satisfy the requirements.
Scenario C (Varying UE Count): As the number of UEs increased (up to 60), EMTAD successfully covered all users and met traffic demands using at most two UAVs. In contrast, the benchmark approach required up to six UAVs for the same scenario without offering significant throughput improvements.

Significance and Claims
The paper claims that EMTAD significantly improves network performance while reducing deployment overhead. By minimizing the number of required UAVs in dynamic environments, the framework ensures efficient resource utilization and reduced operational costs. The authors emphasize that their approach enhances LoS coverage, balances network load, and improves spectral efficiency compared to conventional static strategies. The work builds upon prior research in single-UAV positioning but extends it to address the complexities of multi-UAV optimization and dynamic user association. Future work is identified as extending the framework to support UAV mobility, energy-aware operations, and integration with 6G-native edge intelligence.