WiFo-CF: Wireless Foundation Model for CSI Feedback

TL;DR

WiFo-CF is a versatile wireless foundation model for CSI feedback that uses self-supervised pre-training and a Mixture of Experts architecture to handle diverse configurations and improve generalization across various scenarios.

Contribution

The paper introduces WiFo-CF, a novel foundation model for CSI feedback that supports heterogeneous configurations through innovative pre-training and architecture design.

Findings

Achieves superior performance on diverse datasets

Effectively generalizes to out-of-distribution data

Facilitates downstream tasks like indoor localization

Abstract

Deep learning-based channel state information (CSI) feedback schemes demonstrate strong compression capabilities but are typically constrained to fixed system configurations, limiting their generalization and flexibility. To address this challenge, WiFo-CF, a novel wireless foundation model tailored for CSI feedback, is proposed, uniquely accommodating heterogeneous configurations such as varying channel dimensions, feedback rates, and data distributions within a unified framework through its key innovations: (1) a multi-user, multi-rate self-supervised pre-training strategy; and (2) a Mixture of Shared and Routed Expert (S-R MoE) architecture. Supporting the large-scale pre-training of WiFo-CF is the first heterogeneous channel feedback dataset, whose diverse patterns enable the model to achieve superior performance on both in-distribution and out-of-distribution data across simulated and real-world scenarios. Furthermore, the learned representations effectively facilitate adaptation to downstream tasks such as CSI-based indoor localization, validating WiFo-CF's scalability and deployment potential.