Decentralized GNSS at Global Scale via Graph-Aware Diffusion Adaptation

TL;DR

This paper introduces a decentralized GNSS system that uses graph-aware diffusion and neural networks to achieve high-precision localization at a global scale, improving scalability and efficiency over traditional centralized methods.

Contribution

It presents a novel decentralized GNSS architecture employing graph-aware diffusion and deep learning to enable scalable, resilient, and accurate satellite navigation without centralized processing hubs.

Findings

Achieves centimeter-level self-localization and network-wide consensus.

Matches the accuracy of centralized GNSS systems.

Outperforms existing decentralized methods in convergence speed and communication efficiency.

Abstract

Network-based Global Navigation Satellite Systems (GNSS) underpin critical infrastructure and autonomous systems, yet typically rely on centralized processing hubs that limit scalability, resilience, and latency. Here we report a global-scale, decentralized GNSS architecture spanning hundreds of ground stations. By modeling the receiver network as a time-varying graph, we employ a deep linear neural network approach to learn topology-aware mixing schedules that optimize information exchange. This enables a gradient tracking diffusion strategy wherein stations execute local inference and exchange succinct messages to achieve two concurrent objectives: centimeter-level self-localization and network-wide consensus on satellite correction products. The consensus products are broadcast to user receivers as corrections, supporting precise point positioning (PPP) and precise point positioning-real-time kinematic (PPP-RTK). Numerical results demonstrate that our method matches the accuracy of centralized baselines while significantly outperforming existing decentralized methods in convergence speed and communication overhead. By reframing decentralized GNSS as a networked signal processing problem, our results pave the way for integrating decentralized optimization, consensus-based inference, and graph-aware learning as effective tools in operational satellite navigation.