Structured Latent Dynamics in Wireless CSI via Homomorphic World Models

2026-03-20

AI Summary

This work addresses the challenge of learning structured, geometrically consistent, and generalizable low-dimensional representations from wireless channel state information (CSI). Framing CSI time series modeling as a world model learning problem, we propose a self-supervised latent dynamics model that leverages a Joint Embedding Predictive Architecture (JEPA) conditioned on user actions to predict channel evolution in a compact latent space. A key innovation is the incorporation of a Lie algebra–based homomorphic update mechanism, which inherently encodes spatial layout and user motion into the latent representations, thereby ensuring compositional structure and geometric consistency. Experiments on the DICHASUS dataset demonstrate that our approach significantly outperforms strong baselines, achieving superior performance in preserving topological structure, predicting future embeddings, and generating metrically faithful channel maps—particularly in unseen environments.

Technology Category

• Machine Learning: Learning with Manifolds
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Intelligent Robots: State Estimation

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We introduce a self-supervised framework for learning predictive and structured representations of wireless channels by modeling the temporal evolution of channel state information (CSI) in a compact latent space. Our method casts the problem as a world modeling task and leverages the Joint Embedding Predictive Architecture (JEPA) to learn action-conditioned latent dynamics from CSI trajectories. To promote geometric consistency and compositionality, we parameterize transitions using homomorphic updates derived from Lie algebra, yielding a structured latent space that reflects spatial layout and user motion. Evaluations on the DICHASUS dataset show that our approach outperforms strong baselines in preserving topology and forecasting future embeddings across unseen environments. The resulting latent space enables metrically faithful channel charts, offering a scalable foundation for downstream applications such as mobility-aware scheduling, localization, and wireless scene understanding.

Problem

• Channel State Information
• Latent Dynamics
• World Models
• Wireless Channels
• Structured Representations

Innovation

• homomorphic dynamics
• latent world models
• CSI representation learning
• Lie algebra
• self-supervised wireless modeling