⚡ electrical engineering

Load Balancing in Non-Terrestrial Networks Using Free Space Optical Inter-satellite Links

This paper proposes a fairness-driven load balancing algorithm for Non-Terrestrial Networks that dynamically optimizes ground station selection and utilizes Free Space Optical inter-satellite links to mitigate feeder link bottlenecks, significantly improving worst-case downlink throughput and fairness under both rainy and clear conditions.

Original authors: Abid Afridi, Alexis A. Dowhuszko, Jevgenij Krivochiza, Risto Wichman, Jyri Hämäläinen

Published 2026-03-16

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Abid Afridi, Alexis A. Dowhuszko, Jevgenij Krivochiza, Risto Wichman, Jyri Hämäläinen

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 the Earth is covered in a giant, invisible web of satellites orbiting above us, acting like high-speed cell towers in the sky. These are Non-Terrestrial Networks (NTNs). Their job is to beam internet and data down to people on the ground.

However, there's a problem. Just like how a traffic jam on a highway can leave some drivers stuck while others zoom by, these satellite networks often suffer from uneven traffic.

The Problem: The "Rainy Day" Bottleneck

Think of the connection between a satellite and a ground station (like a big dish antenna on Earth) as a water pipe.

The Good Days: On a clear day, the pipe is wide open, and water (data) flows freely.
The Bad Days: When it rains, the pipe gets clogged. Rain is especially bad for high-frequency signals (like the Ka-band used here), acting like a sponge that soaks up the signal.

In a traditional setup, if a satellite is over a rainy city (like Santiago, Chile, in the paper's example), its pipe gets clogged. That satellite can't send data fast enough, even if the satellites next to it are sitting in sunny skies with wide-open pipes. The network becomes unfair: some users get slow internet, while others get fast internet, even though they are all part of the same system.

The Solution: The "Optical High-Speed Train"

The authors of this paper propose a clever fix: Free Space Optical Inter-Satellite Links (FSO ISLs).

Imagine the satellites aren't just isolated towers; they are connected to each other by invisible laser beams. These are like high-speed, private train tracks running between the satellites in space.

How it works: If Satellite A is stuck in the rain and its pipe to the ground is clogged, it doesn't have to wait. It can instantly shoot its data via a laser beam to Satellite B, which is currently in sunny weather with a clear pipe to the ground. Satellite B then dumps the data onto the ground for everyone.

This is like a delivery driver who hits a traffic jam. Instead of sitting there, they radio a colleague in a nearby car who is on a clear road, hand off the package, and let that colleague finish the delivery.

The "Fairness" Algorithm

The paper introduces a smart computer brain (an algorithm based on Linear Programming) that acts like a traffic controller.

Every few minutes, this controller looks at the whole sky and asks:

Who is in the rain?
Who is in the sun?
Who has the strongest laser connection to whom?

It then solves a giant puzzle to ensure fairness. Instead of trying to make the total amount of data as high as possible (which might still leave one poor satellite behind), it tries to make sure the slowest satellite is as fast as possible. It forces the "rich" satellites (in the sun) to help the "poor" satellites (in the rain) by sharing their capacity.

The Results: Smoother Rides for Everyone

The researchers tested this using a real-world model of the O3b mPOWER satellite system (a fleet of Medium Earth Orbit satellites).

Without the Laser Trains (No ISL): When it rained, some satellites' speeds dropped drastically (like a car hitting a mud puddle). The speeds varied wildly from satellite to satellite.
With the Laser Trains (With ISL): Even when it rained heavily, the system smoothed everything out. The "slowest" satellite didn't slow down much because it could offload its traffic to its neighbors.

The Bottom Line:
By using these laser links between satellites, the network became 25% more reliable during bad weather and 10% better even on sunny days. It didn't need to build more satellites or use more power; it just needed to be smarter about how it shared the load.

In short: The paper teaches us that in space, just like in life, if you want to get through a storm, it helps to have friends standing next to you who can lend you an umbrella (or in this case, a laser beam).

1. Problem Statement

Non-Terrestrial Networks (NTNs), particularly those utilizing Medium Earth Orbit (MEO) constellations, face significant challenges in maintaining uniform downlink performance. The primary issues include:

Uneven Satellite-Gateway Visibility: Satellites experience varying line-of-sight conditions with Ground Stations (GS), leading to some nodes being underutilized while others become bottlenecks.
Weather-Induced Attenuation: Feeder Links (FLs) operating at high frequencies (e.g., Ka-band) suffer severe signal degradation due to rain, causing drastic drops in throughput for specific satellites.
Static Routing Limitations: Traditional routing strategies often optimize for total throughput or shortest paths without considering fairness, leading to "moving bottlenecks" where traffic concentrates on satellites with the best current link conditions, starving others.

The core problem is how to dynamically redistribute traffic across a satellite constellation to maximize the minimum downlink data rate (max-min fairness) for all satellites, especially under adverse weather conditions, without adding new hardware resources.

2. Methodology

The authors propose a fairness-driven load balancing strategy that treats the satellite constellation as a time-varying network. The methodology involves three main components:

A. System Modeling

Topology: A MEO constellation (modeled after the SES O3b mPOWER system) with $K$ satellites and $I$ Ground Stations.
Links:
- Feeder Links (FL): Ka-band (20 GHz) radio frequency links connecting satellites to GSs. Capacity is dynamic, dependent on geometry, off-axis angles, and rain attenuation (modeled using ITU-R P.838-3).
- Inter-Satellite Links (ISL): Free Space Optical (FSO) links (1550 nm) connecting adjacent satellites with fixed, high-capacity, and weather-resilient connections.
Traffic Flow: The system models data flow as an Anycast Multi-Commodity Flow Problem (ACMCFP). Each satellite's data can be delivered to any reachable GS either directly or via a one-hop relay through a neighboring satellite using ISLs.

B. Optimization Formulation

The objective is to maximize the minimum download rate ( $R_k$ ) across all satellites $k$ at every time slot $n$ .

Objective Function: $\max \min_{k \in K} (R_k[n])$
Constraints:
- Flow Conservation: Ensures that incoming traffic from neighbors equals outgoing traffic to GSs or other neighbors.
- Capacity Limits: The sum of allocated fractions for FLs and ISLs cannot exceed physical link capacities.
- Non-negativity: Allocation weights must be valid probabilities ( $0 \leq w, v \leq 1$ ).
Solution Approach: The problem is formulated as a Linear Programming (LP) problem. It is solved at each time slot using the Dual-Simplex algorithm (via MATLAB's linprog) to determine optimal scheduling weights for direct FL transmission and ISL offloading.

C. Algorithm

The proposed algorithm (Algorithm 1) operates iteratively per time slot:

Calculate distances and off-axis angles based on satellite/GS positions.
Compute achievable data rates for FLs (factoring in rain attenuation) and ISLs.
Solve the LP optimization problem to find the optimal traffic split ( $w^*$ for FL, $v^*$ for ISL).
Execute the routing and offloading decisions.

3. Key Contributions

Anycast Multi-Commodity Flow Modeling: The paper uniquely models the MEO constellation as an ACMCFP, allowing data from any satellite to be routed to any available GS, either directly or via ISLs. This flexibility is crucial for load balancing.
Integration of FSO ISLs for Load Balancing: It demonstrates how optical ISLs can act as a "safety valve" to offload traffic from satellites suffering from rain-induced FL degradation to neighbors with clear skies.
Max-Min Fairness Optimization: Unlike previous works focusing on total throughput, this approach explicitly targets the worst-case user/satellite, ensuring a baseline level of service for the entire constellation.
Realistic Simulation Framework: The study uses real TLE (Two-Line Element) data for the SES O3b mPOWER constellation and realistic rain attenuation models for specific global locations (Santiago, Dubbo, Phoenix).

4. Simulation Results

The authors evaluated the system using a 6-satellite MEO constellation and 8 global Ground Stations under various weather conditions.

No-Rain Conditions:
- Without ISL: Throughput varied significantly (625–735 Mbps) with a standard deviation of ~26 Mbps.
- With ISL: Throughput distribution collapsed into a narrow peak around 686 Mbps, reducing the standard deviation by 82% (to ~4.6 Mbps). The average rate remained similar, but fairness improved drastically.
Rain Conditions:
- Without ISL: Heavy rain at specific locations (e.g., Santiago) caused throughput to drop to ~545 Mbps. Moderate rain caused drops to ~583 Mbps.
- With ISL: The algorithm successfully rerouted traffic from rain-affected satellites to neighbors via FSO links.
- Performance Gains:
  - Heavy Rain: Minimum downlink data rate improved by >25%.
  - No Rain: Minimum downlink data rate improved by >10%.
- The ISL-assisted approach maintained a stable minimum throughput of ~680–690 Mbps throughout the 24-hour simulation, effectively mitigating localized feeder-link failures.

5. Significance

This research provides a scalable and practical framework for future 6G and integrated space-ground networks.

Resilience: It proves that optical ISLs are not just for high-speed backhaul but are critical for network resilience against weather-induced outages.
Resource Efficiency: It achieves significant fairness improvements without requiring additional satellites or ground stations, simply by optimizing the routing logic.
Standardization Potential: The proposed linear programming approach offers a computationally feasible method for real-time resource allocation in dynamic NTNs, addressing the "bottleneck" problem inherent in current satellite internet architectures.

In conclusion, the paper demonstrates that combining Free Space Optical ISLs with max-min fairness optimization effectively transforms a vulnerable satellite constellation into a robust, self-healing network capable of maintaining high-quality service even during severe weather events.