Impact of 5G Latency and Jitter on TAS Scheduling in a 5G-TSN Network: An Empirical Study

This empirical study demonstrates that maintaining end-to-end determinism in 5G-TSN networks for IIoT applications requires carefully tuning Time-Aware Shaper (TAS) transmission window offsets based on high-order percentile bounds of 5G downlink delay and jitter to prevent excessive latency or scheduling failures.

The Gist

Here is an explanation of the paper using simple language and creative analogies.

The Big Picture: The "Factory on the Move" Problem

Imagine a high-tech factory where robots and machines need to talk to each other instantly to avoid crashing or making mistakes. This is the Industrial Internet of Things (IIoT).

In the past, these machines were connected by wires (like a long, super-fast train track). This was great because the train always arrived exactly on time. This system is called TSN (Time-Sensitive Networking). It uses a strict schedule called TAS (Time-Aware Shaper) to make sure every packet of data gets a specific "time slot" to travel, like a train leaving a station at exactly 10:00:00 AM.

The Problem:
Now, factories want Autonomous Mobile Robots (AMRs) that can drive around freely. You can't run a wire to a moving robot. So, they use 5G (wireless) instead of wires.

The Conflict:

• TSN (Wired) is like a Swiss watch: precise, predictable, and on time.
• 5G (Wireless) is like a busy city street: sometimes there's traffic, sometimes a red light takes longer than usual, and sometimes a bus blocks the lane. It's unpredictable.

If you try to run a Swiss-watch schedule on a busy city street, the train might arrive late, miss its slot, and cause a pile-up. This paper asks: "How do we adjust the Swiss watch so it still works perfectly, even when the road is a busy city street?"

---

The Core Concept: The "Bus Stop" Analogy

To understand the solution, imagine a bus system.

1. The Master Station (TSN Switch 1): This is the main bus depot. It has a strict schedule. Every 30 minutes, a bus leaves with a group of passengers (data packets).
2. The 5G Highway: This is the road between the depot and the next stop. It's bumpy and unpredictable. Sometimes the bus gets stuck in traffic (delay); sometimes it hits a pothole (jitter).
3. The Slave Station (TSN Switch 2): This is the destination bus stop. It also has a strict schedule. It opens its doors for 5 minutes to let passengers off, then closes them immediately to let the next bus in.

The Challenge:
If the bus leaves the Master Station at 10:00 AM, but gets stuck in traffic and arrives at the Slave Station at 10:15 AM, but the Slave Station only opens its doors at 10:05 AM... The passengers miss the bus! They have to wait for the next bus (the next time slot), which ruins the schedule.

The Solution: The "Buffer Zone" (Offset)

The paper's main discovery is about how to set the Offset (the time difference between when the Master Station opens and when the Slave Station opens).

The "Too Early" Trap (Scenario 3 & 4)

If you set the Slave Station to open its doors only 5 minutes after the Master Station leaves, but the 5G traffic is bad and takes 15 minutes, the passengers arrive late. They miss the door.

• Result: Chaos. Some passengers get on the next bus. The system loses its "determinism" (predictability).

The "Too Late" Trap (Scenario 1 & 2)

If you set the Slave Station to open its doors 60 minutes after the Master Station leaves, the passengers arrive at 10:15 AM, but the doors don't open until 11:00 AM.

• Result: The passengers are safe, but they are sitting in the waiting room for 45 minutes. This adds unnecessary delay.

The "Goldilocks" Solution

The paper says you need to find the perfect middle ground.

1. Measure the Traffic: First, you need to watch the 5G road for a while. You find out that 99.9% of the time, the bus arrives within 15 minutes. (This is the "99.9th percentile").
2. Set the Buffer: You set the Slave Station to open its doors 20 minutes after the Master Station leaves.
   • This gives a 5-minute safety buffer for the rare, super-bad traffic days.
   • It ensures the doors are open before the last passenger arrives.
   • It doesn't wait too long, so the delay stays low.

The "Jitter" Rule (The Size of the Window)

The paper also found a rule about the size of the time window (how long the doors stay open).

Imagine the 5G road is so bumpy that the arrival time of the bus varies wildly (this is Jitter).

• If the "bumpiness" (Jitter) is 10 minutes, but your time window (doors open) is only 5 minutes long, you can't guarantee everyone gets on.
• The Rule: The time between the start of one bus cycle and the start of the next must be longer than the "bumpiness" of the road. If the road is too bumpy, you need to space the buses further apart to avoid collisions.

What Happens When Things Get Crowded?

The researchers also tested what happens when:

1. More Robots (Data Flows): If you have 7 robots talking at once, the 5G road gets more crowded. The "traffic jams" get longer. You have to increase your safety buffer (Offset) even more.
2. Bad Traffic (Best Effort Data): If people are downloading huge videos (Best Effort traffic) on the same 5G network, they clog the road. Even though the robots have "priority," the road gets so full that the robots get delayed. You have to adjust your schedule again to handle this extra weight.

The Takeaway

This paper is like a manual for driving a race car on a bumpy road.

• Old Way: Drive at a fixed speed, hoping the road is smooth. (This fails in 5G).
• New Way (This Paper): Measure how bumpy the road actually is. Then, adjust your pit-stop timing (the Offset) so that you arrive at the pit just as the mechanic is ready, no matter if you hit a pothole or not.

In short: To make 5G work for critical factory robots, you can't just guess the timing. You must measure the worst-case delays and add a smart safety buffer to your schedule. If you do this, you get the speed of 5G with the reliability of a wired connection.

Technical Summary

Here is a detailed technical summary of the paper "Impact of 5G Latency and Jitter on TAS Scheduling in a 5G-TSN Network: An Empirical Study."

1. Problem Statement

The integration of Time-Sensitive Networking (TSN) with 5G private networks is a promising solution for Industrial Internet of Things (IIoT) applications requiring mobility and deterministic communication (e.g., Autonomous Mobile Robots). TSN relies on the IEEE 802.1Qbv Time-Aware Shaper (TAS) to schedule traffic in precise time slots, ensuring bounded latency and low jitter.

However, a critical conflict exists:

• TSN Determinism: Requires strict, synchronized timing across switches.
• 5G Stochasticity: Introduces variable delay and jitter due to radio conditions, scheduling, and queuing in the 5G core and radio access network.

This variability disrupts the synchronized transmission windows of TAS. If the 5G delay causes packets to arrive at the receiving TSN switch outside its allocated transmission window, Inter-Cycle Interference (ICI) occurs. This leads to packet deferral to the next cycle, increased jitter, and a complete loss of end-to-end (E2E) determinism. Existing literature often relies on idealized simulations, lacking empirical validation under commercial 5G conditions to determine how to tune TAS parameters (specifically the offset between switches) to compensate for real-world 5G dynamics.

2. Methodology

The authors conducted a comprehensive empirical study using a custom-built testbed that integrates commercial 5G equipment with industrial TSN switches.

Testbed Architecture:

• 5G System: A private 5G network using Amarisoft SDR (gNB and Core) operating in the n78 band (50 MHz). Two User Equipments (UEs) were connected via Quectel modems.
• TSN Network: Implemented using Safran WR-Z16 switches. The topology included a Master Switch (MS), a Slave Switch (SL), and TSN Translators (NW-TT and DS-TT) to bridge the 5G system.
• Synchronization: High-precision time synchronization was achieved using a Grand Master (GM) clock and PTP (Precision Time Protocol) in Transparent Clock (TC) mode across the 5G translators.
• Traffic: The study focused on a Delay-Critical (DC) flow (simulating Cyclic-Synchronous industrial traffic) and a Best-Effort (BE) flow to test isolation.

Experimental Approach:
The authors performed five distinct experiments to analyze the interaction between 5G delay/jitter and TAS parameters:

1. Delay Characterization: Measured the empirical delay distribution of the 5G system under varying traffic loads to establish statistical bounds (percentiles).
2. Offset Analysis: Varied the time offset ($\delta_{DC}$) between the network cycles of the MS and SL switches to identify the minimum offset required to prevent ICI.
3. Network Cycle Duration: Analyzed the impact of varying the network cycle period ($T_{nc}$) relative to the 5G jitter.
4. Multi-Flow Priority: Evaluated the impact of multiple concurrent DC flows sharing the same priority queue on delay and jitter.
5. BE Traffic Load: Tested the effect of high Best-Effort traffic loads on the latency of the critical DC flow.

3. Key Contributions

The paper makes three primary contributions:

• C1: Analytical Model of 5G-TSN Delay: The authors provide a detailed mathematical model decomposing E2E delay, explicitly characterizing how 5G-induced jitter interacts with TAS parameters (transmission window size, network cycle, and offset). They define the "Zero-Wait-at-SL" (ZWSL) empirical delay as a key metric.
• C2: Conditions for Determinism: They derived specific mathematical conditions required to guarantee deterministic transmission in 5G-TSN networks. Crucially, they identified that the network cycle duration minus the transmission window must be strictly greater than the 5G-induced jitter ($T_{nc} - W_{DC} > t_{jit}$). They also defined the necessary offset between switches based on high-percentile delay bounds.
• C3: Experimental Validation: Unlike prior simulation-based works, this study validates these theories on a real-world commercial testbed. They identified four distinct operational scenarios (ranging from deterministic with early arrival to non-deterministic with ICI) and provided configuration guidelines for each.

4. Key Results

• Dominance of 5G Delay: The 5G system delay (milliseconds) significantly outweighs wired link delays (microseconds). The 99.9th percentile delay ($\hat{D}_{DC, 0.999}$) was found to be approximately 15 ms under moderate load, compared to a wired max of ~40 $\mu$s.
• Offset Configuration: To guarantee determinism, the offset ($\delta_{DC}$) between the MS and SL network cycles must be set to at least the 99.9th percentile of the measured 5G delay.
  • Insufficient Offset: Causes packets to miss the transmission window, leading to ICI and a bimodal delay distribution (packets delayed by a full network cycle).
  • Excessive Offset: Increases E2E latency unnecessarily but maintains determinism.
• Network Cycle Constraint: The study confirmed that if the network cycle is too short (specifically if $T_{nc} - W_{DC} < t_{jit}$), ICI becomes unavoidable regardless of the offset setting. In the experiments, cycles shorter than 17.5 ms failed to maintain determinism due to high jitter.
• Impact of Traffic Load:
  • Same-Priority Flows: Increasing the number of concurrent DC flows significantly increased jitter and the 99.9th percentile delay (up to 22 ms), requiring larger offsets and transmission windows.
  • Best-Effort Traffic: High BE traffic loads (up to 980 Mbps) degraded DC flow performance due to buffer contention in the 5G system, even with QoS mapping. This demonstrated that 5G QoS (5QI) provides relative prioritization but not absolute isolation, necessitating dynamic re-tuning of TAS parameters under heavy load.

5. Significance

This paper is significant for the deployment of Industry 4.0 and CPS (Cyber-Physical Systems) because:

• Bridging Theory and Practice: It moves beyond theoretical models and idealized simulations to provide actionable, data-driven guidelines for configuring TSN over real 5G networks.
• Preventing Determinism Failure: It explicitly identifies the "Inter-Cycle Interference" phenomenon, which is the primary cause of failure in 5G-TSN integrations, and provides the mathematical thresholds to avoid it.
• Configuration Guidelines: It offers a clear methodology for network engineers:
  1. Measure the 5G delay distribution under expected load.
  2. Select a high-percentile (e.g., 99.9%) as the delay bound.
  3. Set the switch offset ($\delta$) to this bound.
  4. Ensure the network cycle is long enough to accommodate the jitter ($T_{nc} > W_{DC} + t_{jit}$).
• Future-Proofing: The findings highlight the need for adaptive algorithms that can dynamically adjust TAS parameters based on real-time 5G channel conditions and traffic loads, a critical step toward robust 6G industrial networks.

In conclusion, the paper establishes that while 5G-TSN integration is feasible, achieving E2E determinism requires careful, empirically derived tuning of TAS parameters to absorb the inherent stochasticity of the wireless medium.