Sensing for Free: Learn to Localize More Sources than Antennas without Pilots

TL;DR

This work tackles pilot-free, multi-source DOA localization during uplink data transmission by reusing unknown data symbols and leveraging sparse MIMO arrays to form a large virtual aperture. It introduces Snap-TF, an attention-only transformer that processes raw SLA snapshots in a grid-less, end-to-end manner to estimate DOAs, effectively capturing higher-order statistics without covariance reconstruction. System-level benefits include sensing-assisted MU-MIMO beam management, enabling significant pruning of downlink beam sweeping and throughput gains, while achieving over a 30x reduction in parameters and runtime compared to state-of-the-art baselines. The approach integrates smoothly with 3GPP NR and 6G standards, enabling pilot-free sensing during regular transmissions and motivating further exploration into privacy, robustness, and wideband/near-field extensions for next-generation networks.

Abstract

Integrated sensing and communication (ISAC) represents a key paradigm for future wireless networks. However, existing approaches require waveform modifications, dedicated pilots, or overhead that complicates standards integration. We propose sensing for free - performing multi-source localization without pilots by reusing uplink data symbols, making sensing occur during transmission and directly compatible with 3GPP 5G NR and 6G specifications. With ever-increasing devices in dense 6G networks, this approach is particularly compelling when combined with sparse arrays, which can localize more sources than uniform arrays via an enlarged virtual array. Existing pilot-free multi-source localization algorithms first reconstruct an extended covariance matrix and apply subspace methods, incurring cubic complexity and limited to second-order statistics. Performance degrades under non-Gaussian data symbols and few snapshots, and higher-order statistics remain unexploited. We address these challenges with an attention-only transformer that directly processes raw signal snapshots for grid-less end-to-end direction-of-arrival (DOA) estimation. The model efficiently captures higher-order statistics while being permutation-invariant and adaptive to varying snapshot counts. Our algorithm greatly outperforms state-of-the-art AI-based benchmarks with over 30x reduction in parameters and runtime, and enjoys excellent generalization under practical mismatches. Applied to multi-user MIMO beam training, our algorithm can localize uplink DOAs of multiple users during data transmission. Through angular reciprocity, estimated uplink DOAs prune downlink beam sweeping candidates and improve throughput via sensing-assisted beam management. This work shows how reusing existing data transmission for sensing can enhance both multi-source localization and beam management in 3GPP efforts towards 6G.

Figures (12)

• Figure 1: SLAs and their variants, including but not limited to MIMO systems with antenna selection Molisch2004MIMO and fluid/movable antennas Wong2021FluidZhu2024Modeling.
• Figure 2: System model. In the uplink, there are $K$ sources impinging upon a BS from DOAs $\bm{\theta}=\{\theta_{1},\theta_{2},\ldots,\theta_{K}\}$. The BS aims to localize these sources based on the received random and unknown uplink data payloads.
• Figure 4: Schematic diagram of the proposed Snap-TF algorithm.
• Figure 5: MSE as a function of SNR under both Gaussian and non-Gaussian signals with $M=5$ and $N=9$. The number of sources and signal types are listed in the title of each subfigure.
• Figure 6: Generalization performance to different numbers of snapshots $T$ and different modulation types. The model is trained under $T=50$ and $\Delta_{\bm{\theta},\min}=\frac{\pi}{60}$, and directly tested in all the different settings.

Theorems & Definitions (1)

• Remark 1: "sensing for free"