🔢 mathematics

USAM: A Unified Safety-Age metric for Timeliness in Heterogeneous IoT Systems

This paper introduces the Unified Safety-Age Metric (USAM), a novel framework that integrates information freshness, deadline reliability, and deterministic response-time constraints to evaluate timeliness in heterogeneous IoT systems, revealing that in ultra-sparse regimes, safety feasibility is governed by receiver readiness rather than traffic load.

Original authors: Mikael Gidlund

Published 2026-03-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Mikael Gidlund

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 the manager of a massive, high-tech factory. You have thousands of workers (IoT devices) sending you messages. Some messages are just "status updates" (like "the temperature is 20°C"), some are "urgent instructions" (like "stop the conveyor belt"), and some are "life-or-death alarms" (like "the robot arm is about to hit a human").

For a long time, factory managers only cared about how fresh the news was. They asked: "How long ago was the last message sent?" This is what computer scientists call Age of Information (AoI). If the news is 5 seconds old, they thought, "Great, that's fresh enough!"

But here is the problem: In a safety-critical factory, "freshness" isn't enough. You need guarantees. Even if the average message is fresh, what if the worst-case scenario happens? What if a life-saving alarm gets stuck in a line for 10 seconds because the manager was taking a coffee break? In the world of safety, 10 seconds is an eternity.

This paper introduces a new tool called USAM (Unified Safety–Age Metric). Think of it as a Super-Manager who doesn't just check the clock; they check the clock, the safety manual, and the emergency exit plan all at once.

Here is the breakdown of how it works, using simple analogies:

1. The Old Way: The "Freshness" Trap

Imagine a newsstand. The old metric (AoI) only cares if the newspaper is today's edition.

The Flaw: If the newsstand is open 99% of the time, the papers are usually fresh. But if the newsstand closes for 10 minutes right when a hurricane warning arrives, the "freshness" metric might still look okay on average. However, for the people in the hurricane, that 10-minute gap is a disaster.
The Reality: In industrial safety, you can't have "average" safety. You need 100% certainty that the alarm will go off within a specific time (e.g., 5 milliseconds). If it takes 6 milliseconds, the system has failed, even if it was fast 99 times out of 100.

2. The New Way: The USAM "Triple-Check"

The USAM metric is like a three-legged stool. If one leg breaks, the whole stool falls. It checks three things simultaneously:

Freshness: Is the information new? (The "News" leg).
Reliability: Did the message actually get through on time? (The "Delivery" leg).
Deterministic Safety: Is the worst-case delay short enough to save a life? (The "Emergency Brake" leg).

The paper argues that you cannot design a safe system by just looking at the first two legs. You must ensure the third leg is strong enough to hold the weight of a disaster.

3. The "Ultra-Sparse" Surprise

The paper studies a specific type of factory: Massive IoT. This is a place with millions of devices, but at any given second, only a tiny few are talking. It's like a stadium of 100,000 people where only 5 people are whispering at once.

The Counter-Intuitive Discovery:
Usually, we think traffic jams happen because there are too many cars. But in this "ultra-sparse" world, there are almost no cars. So, why do messages get delayed?

The paper found that the delay isn't caused by traffic; it's caused by the gatekeeper.

The Analogy: Imagine a bouncer at a club who only opens the door for 10 seconds every minute to save energy. Even if only one person is waiting, if they arrive when the door is closed, they have to wait 50 seconds.
The Lesson: In these massive systems, the bottleneck isn't how many devices are talking; it's how often the receiver is awake. If the "bouncer" (the receiver) is asleep too often, the system fails, no matter how few messages are coming in.

4. The "Feasibility Boundary"

The authors draw a line in the sand called the Feasibility Boundary.

Below the line: The system is safe. The bouncer is awake enough, and the alarms go off in time.
Above the line: The system is broken. Even if the "average" performance looks good, the worst-case scenario means someone could get hurt.

The paper shows that old metrics (like AoI) are "blind" to this line. They might say, "Hey, the average wait time is great!" while the system is actually sitting right on the edge of a cliff. USAM sees the cliff and screams, "Stop! You can't cross here!"

Summary

The Problem: Old metrics measure how "fresh" data is, but they ignore the strict, non-negotiable time limits required for safety (like stopping a robot before it hits a human).
The Solution: USAM is a new scorecard that combines freshness, reliability, and strict safety guarantees into one number.
The Insight: In massive IoT systems, the biggest risk isn't too much data; it's the receiver "sleeping" too often. To be safe, the infrastructure must stay awake enough to handle the worst-case emergency, not just the average day.

In a nutshell: Don't just ask, "Is the news fresh?" Ask, "If the worst thing happens, will the system react fast enough to save us?" USAM is the tool that answers that question.

1. Problem Statement

Massive Internet-of-Things (IoT) deployments in 6G systems must simultaneously support heterogeneous traffic classes: low-rate monitoring, latency-critical control, and safety-critical communication. These systems often operate in an ultra-sparse regime, where the total number of devices is massive, but only a small fraction ( $\rho$ ) is active at any given time.

The core problem identified is the inadequacy of classical timeliness metrics (such as Age of Information - AoI) for safety-critical cyber-physical systems.

Safety-Blindness: Classical metrics focus on the average freshness or stochastic delay. They may indicate "satisfactory" performance even when deterministic safety constraints (e.g., Safety Functional Response Time - SFRT defined in IEC 61784-3) are violated.
Infrastructure Readiness: In massive IoT, receiver chains often operate intermittently (duty-cycled) to save energy. This introduces activation delays that classical metrics fail to capture as a hard feasibility boundary.
Gap: There is a lack of a unified metric that jointly evaluates information freshness, deadline reliability, and worst-case deterministic safety feasibility.

2. Methodology

The paper proposes a system-level framework to analyze and evaluate timeliness under these constraints.

System Model

Traffic Classes: The system handles four classes: Monitoring ( $M$ ), Slow Control ($SC$), Fast Control ($FC$), and Safety ( $S$ ), with a fixed-priority preemptive scheduling discipline ( $S \succ FC \succ SC \succ M$ ).
Intermittent Service: The gateway receiver operates with a duty cycle $\delta$ . When inactive, packets incur an activation delay $V(\delta)$ .
Ultra-Sparse Regime: The aggregate arrival rate scales with the activity factor $\rho$ ( $\lambda(\rho) = \Theta(\rho)$ ). As $\rho \to 0$ , queueing delays vanish, and system behavior is dominated by infrastructure readiness and deterministic constraints.
Safety Constraints: Safety feasibility requires that the Worst-Case Response Time (WCRT), $B_k(\delta)$ , strictly satisfies $B_k(\delta) \leq SFRT_k$ (probability of violation is zero).

The Unified Safety–Age Metric (USAM)

The authors introduce USAM, a composite metric defined as a weighted geometric aggregation of three normalized components:
$\Psi(\rho) = F(\rho)^{w_1} \cdot R(\rho)^{w_2} \cdot S(\rho)^{w_3}$

Freshness ( $F$ ): Based on the Age of Synchronization (AoS), normalized against a target update age.
Reliability ( $R$ ): Based on the Safe Packet Delivery Probability (SPDP), ensuring deadlines are met with a residual safety factor.
Deterministic Safety ( $S$ ): An exponential function of the ratio between the worst-case response time bound and the reference SFRT. This component captures the hard feasibility boundary.

3. Key Contributions

Unified Metric Formulation: The introduction of USAM, which integrates stochastic freshness/reliability with deterministic safety feasibility into a single architecture-aware measure.
Asymptotic Analysis: Derivation of expressions for USAM in the ultra-sparse regime ( $\rho \to 0$ ). The analysis proves that in this regime, queueing delays become negligible, and system timeliness is governed by the receiver duty cycle and deterministic response-time constraints.
Feasibility Thresholds:
- Derivation of a minimum safe receiver duty cycle ( $\delta_{safe}$ ) that satisfies both queue stability and worst-case safety constraints.
- Identification of a structural transition where classical metrics fail to detect infeasibility, while USAM collapses to zero.
Theoretical Insight: Demonstration that classical metrics (AoI, PAoI) are "safety-blind" because they rely on statistical moments (expectations) which can be low even when the probability of violating a hard safety deadline is non-zero.

4. Results

Numerical evaluations compared USAM against AoI, Peak AoI (PAoI), Value of Information (VoI), and Age of Control (AoC) under varying receiver duty cycles ( $\delta$ ) and traffic loads ( $\rho$ ).

Behavior of Classical Metrics: As receiver availability decreases or traffic load increases, classical metrics degrade smoothly and gradually. They suggest the system is still functional even when approaching or exceeding safety limits.
Behavior of USAM: USAM exhibits a sharp structural transition. As the system approaches the deterministic safety boundary (defined by $\delta_{safe}$ ), the metric collapses rapidly. This reveals a "safety-feasible region" that is invisible to classical metrics.
Ultra-Sparse Regime: In low-traffic scenarios, the system is not limited by congestion (queueing) but by the receiver duty cycle. If $\delta < \delta_{safe}$ , the system is safety-infeasible regardless of how low the traffic load is.
Traffic Mix: The results hold across different traffic compositions (monitoring-dominated vs. safety-intensive), confirming that the safety constraint is the dominant factor in determining the feasible operating region.

5. Significance

Design Paradigm Shift: The paper argues that for massive IoT with safety requirements, infrastructure dimensioning should be based on worst-case safety constraints and receiver readiness, not average traffic load.
Safety-Aware Architecture: It provides a foundational framework for designing communication architectures that explicitly guarantee functional safety (e.g., for industrial automation and autonomous transport) rather than relying on probabilistic latency guarantees (like URLLC) which may not satisfy deterministic standards.
Metric Evolution: It moves the field beyond "freshness" (AoI) toward "feasibility," ensuring that communication systems do not appear optimal while actually violating critical safety bounds.

In conclusion, the paper establishes that in heterogeneous, safety-critical IoT systems, deterministic feasibility is the primary constraint, and USAM is the necessary tool to detect the boundary between safe and unsafe operation, a boundary that traditional timeliness metrics completely miss.