Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning a Network Digital Twin as a Hybrid System

Christos Mavridis, Fernando S. Barbosa, Hamed Farhadi, Karl H. Johansson

TL;DR

The paper tackles the challenge of modeling dynamic multi-cell wireless networks with network digital twins (NDTs) by introducing a hybrid NDT whose modes align with base stations and workspace regions. It develops an online deterministic annealing (ODA) based, prototype-driven learning framework to identify and continuously refine the hybrid NDT, including event-trigger mechanisms for time-dependent drift and cell-specific changes. Empirical results on a real two-cell 5G testbed show faster convergence, lower data usage, and robust adaptation to network changes and transient malfunctions compared to a neural-network baseline. The proposed approach enhances the practicality of NDTs for 6G network management and lays groundwork for integration with communication-aware motion planning, offering a scalable and explainable digital twin paradigm.

Abstract

Network digital twin (NDT) models are virtual models that replicate the behavior of physical communication networks and are considered a key technology component to enable novel features and capabilities in future 6G networks. In this work, we focus on NDTs that model the communication quality properties of a multi-cell, dynamically changing wireless network over a workspace populated with multiple moving users. We propose an NDT modeled as a hybrid system, where each mode corresponds to a different base station and comprises sub-modes that correspond to areas of the workspace with similar network characteristics. The proposed hybrid NDT is identified and continuously improved through an annealing optimization-based learning algorithm, driven by online data measurements collected by the users. The advantages of the proposed hybrid NDT are studied with respect to memory and computational efficiency, data consumption, and the ability to timely adapt to network changes. Finally, we validate the proposed methodology on real experimental data collected from a two-cell 5G testbed.

Learning a Network Digital Twin as a Hybrid System

TL;DR

The paper tackles the challenge of modeling dynamic multi-cell wireless networks with network digital twins (NDTs) by introducing a hybrid NDT whose modes align with base stations and workspace regions. It develops an online deterministic annealing (ODA) based, prototype-driven learning framework to identify and continuously refine the hybrid NDT, including event-trigger mechanisms for time-dependent drift and cell-specific changes. Empirical results on a real two-cell 5G testbed show faster convergence, lower data usage, and robust adaptation to network changes and transient malfunctions compared to a neural-network baseline. The proposed approach enhances the practicality of NDTs for 6G network management and lays groundwork for integration with communication-aware motion planning, offering a scalable and explainable digital twin paradigm.

Abstract

Network digital twin (NDT) models are virtual models that replicate the behavior of physical communication networks and are considered a key technology component to enable novel features and capabilities in future 6G networks. In this work, we focus on NDTs that model the communication quality properties of a multi-cell, dynamically changing wireless network over a workspace populated with multiple moving users. We propose an NDT modeled as a hybrid system, where each mode corresponds to a different base station and comprises sub-modes that correspond to areas of the workspace with similar network characteristics. The proposed hybrid NDT is identified and continuously improved through an annealing optimization-based learning algorithm, driven by online data measurements collected by the users. The advantages of the proposed hybrid NDT are studied with respect to memory and computational efficiency, data consumption, and the ability to timely adapt to network changes. Finally, we validate the proposed methodology on real experimental data collected from a two-cell 5G testbed.

Paper Structure

This paper contains 22 sections, 1 theorem, 28 equations, 7 figures.

Table of Contents

  1. Introduction
  2. Contribution
  3. Notation
  4. Hybrid Network Digital Twin Model
  5. Network Properties
  6. Signal Strength:
  7. Latency and Channel Throughput:
  8. Handover Delays:
  9. Network Digital Twin
  10. Network Digital Twin as a Hybrid System
  11. Data-based Hybrid NDT Identification
  12. Prototype-based Learning Model
  13. Online Deterministic Annealing
  14. Time Dependency
  15. Regression Error Event Trigger
  16. ...and 7 more sections

Key Result

Lemma 1

The sequence $\mu_i(n)$ constructed by the recursive updates where $z_t\sim Z$, $s_i:=\mathds{1}_{\left[ \bar{c}_i=c\right]}$, $\sum_t \beta(t) = \infty$, $\sum_t \beta^2(t) < \infty$, and the quantities $\hat{p}(\mu_i|z_t)$ and $\mu_i(t)$ are recursively updated as follows: converges almost surely to a solution of (eq:mu_star).

Figures (7)

  • Figure 1: Illustration of a 2D workspace with two cells and one moving UE providing network measurements.
  • Figure 2: Prototype-based representation of the Hybrid NDT.
  • Figure 3: Experimental apparatus at Ericsson's headquarters, Stockholm, Sweden.
  • Figure 4: Comparison of the training behavior of the proposed hybrid NDT with a neural network regressor.
  • Figure 5: Mode identification error of the hybrid NDT.
  • ...and 2 more figures

Theorems & Definitions (2)

  • Definition 1: Network Digital Twin
  • Lemma 1: mavridis2023online