Space-Air-Ground Integrated Networks

Satellite-Terrestrial Integration with RIS and Digital Twins

This page summarizes the paper "Coverage Optimization in RIS-enabled Satellite-Terrestrial Networks: A Digital Twin-based Spatial-Temporal Approach." The work addresses how LEO satellite systems, terrestrial users, and reconfigurable intelligent surfaces can be jointly optimized under fast orbital dynamics, weather-dependent attenuation, interference coupling, and partial observability.

LEO link RIS reflection

Core Idea

RIS-Assisted Coverage

Direct LEO downlink and RIS-reflected paths are combined through a spatial-temporal coverage-grain model, allowing coverage to account for visibility windows, path loss, weather attenuation, and reflected signal gain.

Digital Twin Synchronization

The digital twin synchronizes scenario parameters such as ephemeris, channel states, RIS phases, user mobility, and interference conditions, creating a high-fidelity optimization environment.

Constrained Learning

A Dec-POMDP and CTDE learning structure enables decentralized satellite decisions while respecting power budgets, RIS stability, collision avoidance, handover continuity, and interference constraints.

Paper Synthesis

Coverage Optimization in RIS-enabled Satellite-Terrestrial Networks: A Digital Twin-based Spatial-Temporal Approach

2026 IEEE Transactions on Cognitive Communications and Networking DOI: 10.1109/TCCN.2026.3678233

The paper proposes a digital twin-based spatial-temporal approach, called DTST, for RIS-enabled satellite-terrestrial networks. Its goal is to maximize time-averaged coverage probability while guaranteeing operational constraints in dynamic LEO environments. The model jointly controls RIS phase configuration and satellite power allocation under partial observability and sim-to-real mismatch.

  • Builds a coverage model that unifies direct satellite paths and RIS-reflected paths using Boolean union operations.
  • Models dynamic factors including LEO geometry, weather attenuation, line-of-sight windows, RIS coupling, interference, and user mobility.
  • Uses centralized training with decentralized execution based on MAPPO for decentralized satellite policy learning.
  • Introduces conservative value learning and rare-event replay to reduce overestimation and improve resilience under simulation-to-reality mismatch.
  • Reports more than 80% coverage probability and over 90% prediction accuracy in a Ka-band 48-LEO-satellite constellation under heavy rainfall and co-channel interference.

Optimization Pipeline

1. Sense Scenario

Collect satellite ephemeris, GBU state, weather attenuation, RIS phase state, and channel conditions.

2. Build DT State

Synchronize the physical network with a spatial-temporal digital twin for coverage prediction.

3. Learn Policy

Train decentralized satellite policies through CTDE and MAPPO under partial observability.

4. Calibrate Risk

Use conservative value learning and rare-event replay to reduce value bias and constraint violations.

5. Control Coverage

Jointly adjust satellite power and RIS phases for service continuity and robust coverage.

Key Metrics and Constraints

Coverage Probability

Balances the probability of service availability for ground users against power and interference constraints.

Operational Safety

Accounts for power budgets, RIS stability, LOS windows, collision avoidance, handover continuity, and interference coupling.

Sim-to-Real Robustness

Mitigates digital-twin model mismatch through conservative value calibration and rare-event replay.

Research Implications

From Static Coverage to Spatial-Temporal Network Intelligence

This work reframes satellite-terrestrial integration as a dynamic, digital-twin-driven control problem. Instead of optimizing coverage from a static link budget alone, the DTST framework treats orbital motion, weather, RIS reflection, interference, and handover continuity as coupled spatial-temporal factors. This is directly relevant to future space-air-ground networks that must provide resilient connectivity for urban canyons, mobile field units, IoT/IoV services, and emergency communication.

  • Architecture: RIS adds controllable passive aperture gain to compensate for blocked or weakened LEO links.
  • Control: DT-synchronized learning enables proactive coverage management rather than reactive adjustment.
  • Reliability: rare-event replay and conservative value learning improve stability under severe attenuation and co-channel interference.
Back to Space-Air-Ground Integrated Networks