MultiPathFormer: Towards a Foundation Model for Multipath Wireless Propagation

2026-08-05

AI Summary

This work addresses the limitations of existing wireless foundation models, which struggle to effectively model multipath effects due to their neglect of underlying propagation physics. To overcome this, the authors propose treating multipath propagation as the fundamental pretraining unit, representing transmitter–receiver links as continuous sequences of path tokens and predicting subsequent paths in an autoregressive manner. Built upon a Transformer architecture, the method innovatively integrates an environment-augmented retrieval (RAG) mechanism and a first-path codebook, enabling, for the first time, transferable path-level representations of wireless propagation. The approach significantly outperforms current models across multiple tasks—including localization, beam prediction, line-of-sight classification, and channel estimation—achieving a localization error as low as 5.57 meters and a Top-3 beam prediction accuracy of 0.914.

Abstract

Recent advances in machine learning have enabled training of wireless foundation models, which aim to support tasks such as channel estimation, beam prediction, and localization based on wireless signals. Existing wireless foundation models typically pretrain on channel tensors using masked reconstruction over subcarriers, antennas, or time but ignore the physical characteristics of wireless propagation. In this work, we propose to instead use multipath propagation as the fundamental pretraining object. We present MultiPathFormer, an autoregressive foundation model that represents each transmitter-receiver link as an ordered sequence of continuous-valued path tokens and pretrains with next-path prediction. We introduce an Environmental RAG (retrieval-augmented generation) mechanism and a first-path codebook on top of the transformer backbone, leveraging environment knowledge to improve path statistics estimation like delay and power by up to 59%. MultiPathFormer pretrained on 27 environments transfers to unseen users and, after scenario-specific fine-tuning, outperforms training the corresponding models from scratch in new environments. Across downstream tasks, it outperforms SOTA channel-based foundation models, achieving 5.57 m mean localization error, 0.914 top-3 beam accuracy, 0.994 line-of-sight classification accuracy, and 0.561 channel estimation NMSE. These results show that path-level pretraining can learn reusable representations of wireless propagation.

Problem

• multipath propagation
• wireless foundation model
• channel estimation
• beam prediction
• localization

Innovation

• MultiPathFormer
• multipath propagation
• foundation model
• retrieval-augmented generation
• path-level pretraining