Cross-domain Learning Framework for Tracking Users in RIS-aided Multi-band ISAC Systems with Sparse Labeled Data

TL;DR

This work tackles UE tracking in RIS‑aided multi‑band ISAC systems using CSI indicators available from UL transmissions. It introduces X$^2$Track, a hierarchical, transformer‑based tracking function augmented with adversarial cross‑domain learning to mitigate sparse target‑domain labels by leveraging abundant source‑domain data. The framework encodes multi‑modal frame‑level CSI from direct and RIS paths, aggregates them with a time‑aware sequence encoder, and regresses 3D positions via a lightweight MLP, while aligning source and target feature spaces through a domain classifier trained with a gradient reversal. Simulation results show decimeter‑level axial tracking accuracy under strong interference and scarce labeling, and strong generalization to diverse target domains with less than 5% labeled data, highlighting practical viability for scalable, cross‑domain RIS‑MB ISAC deployments.

Abstract

Integrated sensing and communications (ISAC) is pivotal for 6G communications and is boosted by the rapid development of reconfigurable intelligent surfaces (RISs). Using the channel state information (CSI) across multiple frequency bands, RIS-aided multi-band ISAC systems can potentially track users' positions with high precision. Though tracking with CSI is desirable as no communication overheads are incurred, it faces challenges due to the multi-modalities of CSI samples, irregular and asynchronous data traffic, and sparse labeled data for learning the tracking function. This paper proposes the X2Track framework, where we model the tracking function by a hierarchical architecture, jointly utilizing multi-modal CSI indicators across multiple bands, and optimize it in a cross-domain manner, tackling the sparsity of labeled data for the target deployment environment (namely, target domain) by adapting the knowledge learned from another environment (namely, source domain). Under X2Track, we design an efficient deep learning algorithm to minimize tracking errors, based on transformer neural networks and adversarial learning techniques. Simulation results verify that X2Track achieves decimeter-level axial tracking errors even under scarce UL data traffic and strong interference conditions and can adapt to diverse deployment environments with fewer than 5% training data, or equivalently, 5 minutes of UE tracks, being labeled.

Figures (10)

• Figure 1: An RIS-aided MB ISAC system and the wireless channel for the UL transmission from a UE to the BS in Band $n$.
• Figure 2: Illustration of the RIS-aided MB ISAC protocol, including the components of MB CSI sequnces and the estimation of UE track.
• Figure 3: Transformer-based tracking function $\bm f$ based on the hierarchical architecture proposed in the X$^2$Track framework.
• Figure 4: Average axial tracking errors of $\bm f$ for training and test sets versus the number of training epochs, given different training algorithms.
• Figure 5: (a) Avg. and Std. of axial tracking errors for using CSI indicators of different modalities and different bands; (b) Horizontal distributions of axial tracking errors for using CSI indicators of different bands under the same measurement noise power and the same SNR of CSI indicator conditions.