Fusing Digital Twins and World Models for Proactive Risk-Aware Routing in 6G Satellite Networks

Abstract

For applications such as global broadband access, emergency communications, and integrated space-air-ground networking, 6G networks impose requirements on satellite communication systems. These requirements include high reliability, low latency, and proactive intelligent control. However, low Earth orbit (LEO) satellite networks have highly time-varying topologies. Wireless optical intersatellite links (WOISLs) are also vulnerable to space debris blockage and sun outage. Conventional reactive routing schemes rely on current network states and are difficult to cope with abrupt future changes in link risk. To address this issue, this paper proposes Digital-Twin and World-Model driven Risk-Aware Routing (DT-WM-RAR), a proactive risk-aware routing framework that fuses digital twins (DTs) and world models (WMs). First, a multi-source space risk model is developed for WOISLs. It converts space debris blockage and sun outage into unified link-level risk features. Then, a DT platform for 6G LEO satellite networks is constructed to synchronize constellation topology, link performance, traffic load, and space-environment risk states in real time. On this basis, a latent-space WM is introduced to infer short-term future changes in link risk, load, and connectivity. A Soft Actor-Critic (SAC) policy network is used to generate risk-aware link costs. The service forwarding path is finally obtained through shortest-path search. Simulation results show that the proposed DT-WM-RAR improves the service success rate, reduces rerouting frequency, and decreases end-to-end delay in both low-risk and high-risk scenarios. In particular, under high-risk and heavy-load conditions, it better avoids potentially failed links and improves the reliability and stability of 6G satellite networks in complex space environments.