The idea of using drones as cell towers is not new. We’ve covered this in disaster response scenarios, from drones building temporary mobile networks for mountain rescue teams in Wales to Nokia’s AI-powered drone network for emergency services in Switzerland. However, the fundamental problem still remains. How do you know exactly how many drones to deploy and where to place them without wasting energy or putting users out of range?
team of Illinois Institute of Technology (Illinois Institute of Technology) has found the answer. This includes training machine learning algorithms to make deployment decisions in near real-time.
- development: Professor of Electrical and Computer Engineering, Illinois Institute of Technology Yu Chen and his students published papers IEEE Transactions on Vehicle Technology We describe a machine learning method that jointly optimizes both the number and placement of UAV base stations to supplement fixed cell towers.
- “so what?”: The algorithm can calculate optimal drone deployments fast enough to adjust network coverage over the course of a day, not just weeks or months.
- source: Illinois Institute of Technology news release, published February 10, 2026.
Landline cell towers waste resources that could be recovered by drones
Today’s network coverage relies on fixed base stations, large towers that consume large amounts of power and provide coverage to set geographic zones. These towers are expensive, fixed, and built to handle the maximum expected capacity in the area. If demand drops at 2 a.m. on Tuesday, that excess capacity will remain idle. When demand spikes during large events, towers are less flexible to accommodate it.
Introducing drones as auxiliary base stations changes this equation. UAVs mounted on rooftops or streetlights can extend network bandwidth to areas of high demand and be stored when not needed. The technology to achieve this already exists. Mathematics is difficult.
“How many UAVs should we deploy?” Chen asked. “If this number is too large, we waste energy unnecessarily, and if it is too small, we won’t have enough frequency bandwidth to support our users, so we want to find the optimal number.”
Machine learning approaches solve things that traditional methods cannot
The Illinois Institute of Technology’s approach uses machine learning to consider two variables simultaneously: the number of drones deployed and their physical location. This is a combinatorial optimization problem, and traditional methods for solving it rely on computationally time-consuming approximations. This is sufficient for static planning. It is not useful for real-time deployment decisions.
Cheng’s team combined two machine learning techniques to build an algorithm that quickly arrived at the optimal solution. The algorithm was trained on the results of these slower traditional methods and learned to reproduce their accuracy in a fraction of the processing time.
“Our method has the potential to be a very powerful solution for this category of combinatorial optimization problems,” Cheng said.
Fast calculations allow network providers to adjust drone-based coverage throughout the day as demand changes. Think of this as dynamic load balancing. However, it is targeted at wireless network infrastructure, not web servers.
This research fits into the broader push towards drone-based network infrastructure.
This work does not stand alone. of F.C.C. Shenzhen has been moving to support drone communications for years, proposing new frequency rules for drone operations in 2023 and formally adopting rules for the 5030-5091MHz band in 2024. The city of Shenzhen has committed $1.7 billion to drone infrastructure, including more than 8,000 5G advanced base stations for low-altitude drone operations. Israel has been testing drone connectivity with 5G networks.
The missing piece in most of these efforts is an optimization layer. You can build spectrum rules and 5G towers, but figuring out where to actually place mobile airborne nodes in real time is a whole different ball game. That’s what the Illinois Institute of Technology research is addressing.
Cheng expects some version of drone-based network coverage to become a reality within five years. “This is similar to having UAVs deliver things. It’s currently being done on a small scale and is still experimental, but in theory it’s all doable,” he said.
Research contributors include Oluwaseun T. Ajayi (MS ECE ’24, Ph.D. EE Candidate) and Suyang Wang (MS EE ’17, Ph.D. EE ’24).
DroneXL opinion
We’ve been covering the concept of ‘drones as mobile phone towers’ at least since 2022, when we reported on the Welsh Mountain Rescue Network project. What has changed is that software is finally catching up with hardware. spectrum is assigned. 5G networks are being built. Drones can fly. But without smart deployment algorithms, you just throw expensive UAVs at the problem and hope for the best.
What I focused on in this research was its speed. Being able to recalculate the optimal placement of the drone throughout the day can transform this from a static infrastructure play to something truly responsive. Imagine a concert venue during harvest season, a natural disaster zone, or a rural area where connectivity demand spikes and dips over short periods of time. Fixed towers cannot accommodate that. A fleet of drone base stations guided by this kind of algorithm can do just that.
The five-year timeline Chen mentioned seems about right. Telecommunications companies such as Nokia already operate drone-as-a-service networks in Switzerland, and the FCC continues to open up spectrum for drone communications, laying the groundwork for regulation. By 2028, we expect to see the first commercial pilot programs for drone-based network complementation in disaster response and relaying large-scale events. The explosion of IoT will further accelerate demand from there.
Editor’s note: AI tools were used to assist in researching this article and searching archives. All reporting, analysis and editorial viewpoints are written by Haye Kesteloo.
Learn more at DroneXL.co
Subscribe to receive the latest posts by email.
