PilotWiMAE: Wireless Channel Pilots Are All You Need

Berkay Guler, Giovanni Geraci, Hamid Jafarkhani

Keywords: self-supervised learning, representation learning, wireless channel representation

TL;DR: PilotWiMAE learns wireless channel representations directly from sparse noisy pilot observations using physics-aligned factorized attention, enabling robust feature learning and practical deployment without full-channel knowledge.

Abstract: Wireless channel foundation models assume access to fully observed channels, an assumption that fails in deployment. We introduce PilotWiMAE, a self-supervised framework whose encoder ingests noisy pilot observations directly and whose attention factorizes along the axis separating temporal from joint space-frequency processing, an inductive bias grounded in the physics of the problem. Pilot input shrinks the observation space by up to two orders of magnitude, while the factorized design yields more robust representations by exploiting separable channel structure and allowing a 99% pretraining mask ratio. Pilot-only processing also removes the unrealistic assumption of full-CSI availability while incurring lower latency. We pair patch-normalized reconstruction, which captures small-scale fading structure, with an auxiliary scale loss that recovers the large-scale fading features, and use an AWGN curriculum to match pilot noise at pretraining and deployment. Pretrained solely on 3.5 GHz and evaluated at 28 GHz across in-distribution and out-of-distribution settings, PilotWiMAE's cross-frequency beam selection and channel characterization beat supervised baselines despite operating on a significantly smaller observation space.

Submission Number: 18