CMANet: Channel-Masked Attention Network for Cooperative Multi-Base-Station 3D Positioning

TL;DR

CMANet tackles sub-meter 3D positioning in urban multipath by fusing raw CSI from multiple base stations through Channel Masked Attention and a frequency-sequence LSTM decoder. It introduces a native CSI fusion paradigm and CMA to weight per‑BS links by channel gains, enabling exploitation of cross‑BS multipath complementarity. In simulations with six BSs in a 5G NR urban setting, CMANet achieves a median error below $0.5\text{ m}$ and a $90^{th}$ percentile error below $1.0\text{ m}$, outperforming fingerprint and self‑attention baselines. The approach is edge‑deployable and aligned with ISAC, offering robust, scalable urban localization for autonomous mobility and urban robotics.

Abstract

Achieving ubiquitous high-accuracy localization is crucial for next-generation wireless systems, yet remains challenging in multipath-rich urban environments. By exploiting the fine-grained multipath characteristics embedded in channel state information (CSI), more reliable and precise localization can be achieved. To address this, we present CMANet, a multi-BS cooperative positioning architecture that performs feature-level fusion of raw CSI using the proposed Channel Masked Attention (CMA) mechanism. The CMA encoder injects a physically grounded prior--per-BS channel gain--into the attention weights, thus emphasizing reliable links and suppressing spurious multipath. A lightweight LSTM decoder then treats subcarriers as a sequence to accumulate frequency-domain evidence into a final 3D position estimate. In a typical 5G NR-compliant urban simulation, CMANet achieves less than 0.5m median error and 1.0m 90th-percentile error, outperforming state-of-the-art benchmarks. Ablations verify the necessity of CMA and frequency accumulation. CMANet is edge-deployable and exemplifies an Integrated Sensing and Communication (ISAC)-aligned, cooperative paradigm for multi-BS CSI positioning.

Figures (6)

• Figure 1: Multi-base stations joint positioning scenario
• Figure 2: The architecture of CMANet based on the channel masked attention
• Figure 3: Positioning Accuracy During Training
• Figure 4: CDF of Different Algorithms
• Figure 5: Cumulative Effect of Frequency Domain Features