Digital-Twin-Aided Dynamic Spectrum Sharing and Resource Management in Integrated Satellite-Terrestrial Networks

The explosive growth in wireless service demand has prompted the evolution of integrated satellite-terrestrial networks (ISTNs) to overcome the limitations of traditional terrestrial networks (TNs) in terms of coverage, spectrum efficiency, and deployment cost. Particularly, leveraging LEO satellites and dynamic spectrum sharing (DSS), ISTNs offer promising solutions but face significant challenges due to diverse terrestrial environments, user and satellite mobility, and long propagation LEO-to-ground distance. To address these challenges, digitial-twin (DT) has emerged as a promising technology to offer virtual replicas of real-world systems, facilitating prediction for resource management. In this work, we study a time-window-based DT-aided DSS framework for ISTNs, enabling joint long-term and short-term resource decisions to reduce system congestion. Based on that, two optimization problems are formulated, which aim to optimize resource management using DT information and to refine obtained solutions with actual real-time information, respectively. To efficiently solve these problems, we proposed algorithms using compressed-sensing-based and successive convex approximation techniques. Simulation results using actual traffic data and the London 3D map demonstrate the superiority in terms of congestion minimization of our proposed algorithms compared to benchmarks. Additionally, it shows the adaptation ability and practical feasibility of our proposed solutions.

💡 Research Summary

**
The paper tackles the pressing challenge of efficient spectrum utilization and resource management in integrated satellite‑terrestrial networks (ISTNs), where low‑Earth‑orbit (LEO) satellites complement terrestrial base stations (TBS) to provide ubiquitous coverage. Traditional approaches to dynamic spectrum sharing (DSS) in such networks have relied on snapshot or purely statistical channel models, which cannot capture the rapid variations caused by satellite mobility, long propagation delays, and complex urban propagation environments. To overcome these limitations, the authors introduce a digital‑twin (DT)‑enabled framework that jointly optimizes long‑term and short‑term resource decisions across multiple time scales.

System Overview
A 5G‑NR downlink scenario is considered, comprising N terrestrial access points (APs) and a single LEO satellite serving K user equipments (UEs). Three service classes are defined: delay‑sensitive (e.g., URLLC) served exclusively by terrestrial APs, satellite‑only communication, and a multi‑network service that can be delivered by either. The total bandwidth W_tot is dynamically partitioned into three bandwidth parts (BWPs) for these services, each employing a distinct numerology (μ_D = 2, μ_M = 1, μ_S = 0) reflecting their latency tolerance. Time is organized into cycles, frames (10 ms), and sub‑frames (1 ms), matching the 5G‑NR frame structure.

Digital Twin Model
The DT continuously ingests real‑system data: UE positions and trajectories (including vehicle routes extracted from Google Navigator), traffic arrival logs, and measured channel state information (CSI). Using a 3‑D map of London and a ray‑tracing engine, the DT reconstructs the physical propagation environment, enabling accurate prediction of future CSI and traffic loads. A compressed‑sensing (CS) module exploits sparsity in the CSI domain to reconstruct high‑dimensional channel vectors from limited measurements, while a prediction engine forecasts the next cycle’s channel and traffic conditions.

Optimization Problems
Two coupled optimization problems are formulated:

1. Joint‑RA (Resource Allocation) – Leveraging DT‑predicted information, this mixed‑integer non‑linear program (MINLP) simultaneously decides:

   • Long‑term traffic steering (how much traffic is routed to the satellite versus terrestrial APs);
   • Bandwidth allocation among the three BWPs;
   • Preliminary short‑term decisions: UE‑AP association, resource‑block (RB) assignment, and transmit power levels. The objective is to minimize system congestion, quantified by the sum of queue lengths across all services, subject to QoS latency constraints, power budgets, and inter‑system interference limits.

2. Refinement – At each sub‑frame, real‑time CSI and actual traffic arrivals are used to adjust the preliminary short‑term decisions. This problem focuses on fine‑tuning power control and RB re‑allocation to align the DT‑based plan with the instantaneous network state.

Both problems are non‑convex. The authors address them with two algorithmic strategies:

• DT‑JointRA – The CS‑reconstructed CSI is first sparsified, then successive convex approximation (SCA) iteratively convexifies the non‑linear constraints, yielding a sequence of tractable convex sub‑problems. Integer variables (association, RB assignment) are handled via relaxation and rounding, while the continuous variables (power) are updated within the SCA loop.

• RT‑Refine – A lightweight Lagrangian dual method updates the dual variables associated with power and interference constraints, enabling rapid convergence (typically 3–5 iterations) within each sub‑frame. This ensures that the refined decisions respect the actual measured CSI.

Complexity analysis shows DT‑JointRA scales polynomially (≈ O(N³)) with the number of APs and UEs, whereas RT‑Refine operates in O(N²) per sub‑frame, making real‑time execution feasible.

Performance Evaluation
Simulations employ real traffic traces from a UK operator and a high‑fidelity 3‑D model of London, capturing building‑induced blockage and multipath effects via ray‑tracing. Benchmarks include static bandwidth partitioning, priority‑based sharing, and a recent deep‑learning‑driven allocation scheme. Results demonstrate that the proposed DT‑enabled algorithms reduce average queue lengths by over 30 % compared with all baselines, with the most pronounced gains during peak traffic periods and in densely built‑up zones where interference is severe. Moreover, the pre‑optimization performed by DT‑JointRA allows RT‑Refine to converge quickly, cutting overall control latency by roughly 40 %.

Contributions and Outlook
The paper’s primary contributions are:

1. A novel DT framework that integrates realistic 3‑D environmental modeling, user mobility, and compressed‑sensing‑based CSI prediction for ISTNs.
2. Formulation of a joint long‑term/short‑term resource allocation problem that captures both traffic steering and fine‑grained radio resource control.
3. Efficient solution methods combining CS reconstruction, SCA, and Lagrangian refinement, with demonstrated polynomial complexity.
4. Extensive validation using real‑world data and a city‑scale 3‑D map, confirming both congestion reduction and practical feasibility.

Future work is suggested on extending the framework to multi‑satellite constellations, incorporating inter‑operator spectrum sharing, and exploring reinforcement‑learning agents that can adaptively tune the DT prediction models and optimization parameters in a fully autonomous manner.