⚡ electrical engineering

Data-Driven Adaptive Resource Allocation for Reliable Low-Latency Uplink Communications in Rural Cellular 5G Multi-Connectivity

This paper evaluates multi-connectivity strategies in commercial 5G networks through field measurements across various environments, ultimately proposing a "Primary-Anchored Adaptive Failover" (PAAF) framework and demonstrating that partial duplication can effectively balance reliability and overhead in coverage-constrained rural scenarios.

Original authors: Carlos S. Alvarez-Merino, Alejandro Ramirez-Arroyo, Rasmus Suhr Mogensen, Morten V. Pedersen, Miguel Villanueva-Fernández, Emil J. Khatib, Sergio Fortes, Raquel Barco, Preben E. Mogensen

Published 2026-04-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Carlos S. Alvarez-Merino, Alejandro Ramirez-Arroyo, Rasmus Suhr Mogensen, Morten V. Pedersen, Miguel Villanueva-Fernández, Emil J. Khatib, Sergio Fortes, Raquel Barco, Preben E. Mogensen

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 you are a remote surgeon performing a delicate operation using a robotic arm, or a technician operating a heavy crane from a safe distance. For these jobs, your "connection" (the internet/cellular signal) isn't just about speed; it’s about reliability and timing. If the signal lags for even half a second, or if a single command is lost, the consequences could be catastrophic.

This paper investigates how to keep that connection rock-solid, especially in "rural" areas where cell towers are far apart and signals are weak.

The Problem: The "Rural Struggle"

Think of a cellular signal like a conversation in a crowded, noisy room.

In the City (Urban): There are many people (cell towers) shouting clearly. Even if one person stops talking, another is right there to pick up the conversation. It’s easy to stay connected.
In the Countryside (Rural): There is only one person talking, and they are standing very far away. To be heard, they have to scream at the top of their lungs (this is "Uplink Power Control"). But eventually, they run out of breath. When they can't scream any louder, the connection starts to stutter, and your commands get lost in the silence.

The researchers found that in these rural areas, simply having "bars" on your phone (RSRP) doesn't tell the whole story. You might have signal, but if the device is "out of breath" (power-limited), the connection will still fail.

The Failed Solutions: "Splitting the Load"

When things get bad, engineers usually try two things:

Link Aggregation (The "Two-Lane Highway"): You split your data between two different towers to make a wider road. The researchers found this doesn't work well in the country. If both towers are on the same hill, and a storm rolls in, both "lanes" get blocked at the same time. It’s like trying to use two different roads that both lead through the same narrow, muddy tunnel.
Switching (The "Relay Race"): You use Tower A, but the moment it gets weak, you jump over to Tower B. The problem? By the time you realize Tower A is failing, you've already missed a few important commands. It's like a runner tripping; by the time they stand up, the race is already lost.

The Hero: The "PAAF" Strategy (The "Smart Backup")

The researchers proposed a new smart system called PAAF (Primary-Anchored Adaptive Failover).

Instead of constantly using two connections (which is expensive and wastes battery) or waiting until things break to switch, PAAF acts like a smart bodyguard.

Imagine you are walking through a dark forest with a flashlight (your primary connection).

Normal Mode: You just use the flashlight. It’s efficient and saves battery.
The "PAAF" Mode: The bodyguard is watching the flashlight. The moment the light starts to flicker or the battery looks low, the bodyguard doesn't wait for you to trip; they immediately turn on a second, backup flashlight (this is called Partial Duplication).

For a short period, you have two lights shining at once. This ensures that even if one light fails, the path is still bright. Once the forest clears and the first light is steady again, the bodyguard turns the backup off to save energy.

The Result: Reliability without the Waste

The researchers tested this "Smart Backup" in the real world. They found that:

Full Duplication (always using two connections) is like carrying two heavy flashlights all the time. It’s very safe, but it's exhausting and wasteful.
PAAF gives you almost the same level of safety as the "double flashlight" method, but it only turns the second one on when it's actually needed.

In short: The paper proves that for critical tasks in remote areas, we shouldn't just try to build "wider roads" or "faster runners." Instead, we need smart systems that can sense when a connection is about to struggle and instantly provide a redundant "safety net" before the failure actually happens.

Technical Summary: Data-Driven Adaptive Resource Allocation for Reliable Low-Latency Uplink Communications in Rural Cellular 5G Multi-Connectivity

1. Problem Statement

While 5G networks have significantly improved downlink (DL) performance, the uplink (UL) remains a bottleneck, particularly in rural deployments. Mission-critical applications (e.g., teleoperation, autonomous navigation, and real-time telemetry) require strict low-latency and high-reliability guarantees. In rural areas, these requirements are often violated due to:

Uplink Power-Limited Operation: User Equipment (UE) frequently reaches its maximum transmit power ( $P_{max,UE}$ ) to compensate for path loss, leaving no headroom for further link adaptation.
Coverage Constraints: Sparse infrastructure leads to severe latency spikes and packet loss.
Inefficiency of Current Strategies: Traditional Full Packet Duplication (FD) provides high reliability but incurs 100% overhead (doubling resource usage), while Link Aggregation and Switching-based diversity often fail in rural settings due to correlated fading (co-located base stations) and slow reaction times to transient impairments.

2. Methodology

The researchers conducted extensive real-world measurement campaigns using a mobile vehicle-mounted testbed equipped with two independent 5G Non-Standalone (NSA) modems (two different Mobile Network Operators, MNOs) to enable Multi-Connectivity (MC) evaluation.

Experimental Scenarios: Three environments were tested: Urban (high density), Suburban (mixed), and Rural (sparse/challenging).
Traffic Model: Constant-bit-rate UDP telemetry (ranging from 4 Mbps down to 0.25 Mbps) was used to simulate video monitoring and teleoperation.
Key Performance Indicators (KPIs): The study monitored RSRP, UL Transmit Power, handover frequency, E2E one-way latency (90th, 95th, 99th percentiles), and packet loss (including "late losses" $>800$ ms).
Proposed Framework (PAAF): The authors introduced the Primary-Anchored Adaptive Failover (PAAF) framework, featuring two modes:
- PAAF Switching: Dynamically selects one active interface based on a joint score of RSRP, UL Tx Pwr, and Latency.
- PAAF Partial Duplication (PD): Selectively activates redundancy (duplication) only when the primary link's KPIs cross specific degradation thresholds.

3. Key Contributions

Empirical Characterization: A large-scale study proving that latency and reliability cannot be predicted by coverage indicators (RSRP) alone, especially in power-limited regimes.
Link-Budget Analysis: A detailed investigation into how reducing the UL data rate affects transmit power and latency, revealing that practical 5G link adaptation does not scale linearly with Shannon’s capacity theorem.
PAAF Algorithm Design: A control-aware framework that uses both open-loop (radio KPIs) and closed-loop (latency) metrics to trigger connectivity changes.
Reliability–Efficiency Modeling: The introduction of a "normalized overhead" model to evaluate the trade-off between the cost of redundancy and the suppression of latency tails.

4. Key Results

Rural Vulnerability: Rural scenarios showed orders-of-magnitude increases in 99th-percentile latency and packet loss compared to urban areas. This is strongly correlated with the UE reaching its maximum transmit power.
Failure of Aggregation: Link aggregation was found ineffective in rural areas because splitting the load does not resolve the underlying power-limited bottleneck.
Superiority of Partial Duplication (PD):
- PAAF PD achieved reliability levels nearly identical to Full Duplication (FD) but with significantly lower overhead (reducing duplication by up to ~75%).
- Radio-based triggers (RSRP/Tx Pwr) proved highly effective for PD, acting as "early warning" systems that activate redundancy before severe latency spikes occur.
Switching Limitations: Switching-based strategies performed poorly when MNO base stations were co-located (common in rural areas), as the links became spatially correlated, meaning a failure on one link often coincided with a failure on the other.

5. Significance

This research provides a blueprint for implementing resilient, cost-effective connectivity for Industrial IoT (IIoT) and autonomous systems in coverage-constrained environments. By moving away from "always-on" duplication toward condition-aware partial duplication, operators and developers can meet the stringent requirements of mission-critical applications without the prohibitive spectral and operational costs of constant redundancy. The study highlights that for 5G UL reliability, power-control awareness is just as critical as signal-strength awareness.