Generative MIMO Beam Map Construction for Location Recovery and Beam Tracking

TL;DR

The paper tackles locating users and constructing MIMO beam maps from sparse, unlabeled CSI by introducing a radio-map-embedded generative framework. It combines a dual-scale CSI feature extractor with a hybrid RNN-CNN encoder to capture mobility, and anchors the latent space with a learnable radio map prior, while a diffusion-based decoder reconstructs full CSI conditioned on positional features. Key contributions include a compact radio-map embedding, a diffusion-based generator, and a composite training objective, yielding localization improvements over 30% and beam-tracking capacity gains over 20% versus Kalman-filter baselines, with robustness to very sparse measurements. The approach is validated on outdoor and indoor datasets, showing strong generalization, reliable radio map reconstruction at low measurement rates, and effective real-time beam tracking, underscoring practical impact for scalable location-aware wireless networking. The framework offers a data-efficient path to high-fidelity channel generation, location recovery, and beam management in dynamic environments.

Abstract

Machine learning (ML) has greatly advanced data-driven channel modeling and resource optimization in wireless communication systems. However, most existing ML-based methods rely on large, accurately labeled datasets with location information, which are often difficult and costly to obtain. This paper proposes a generative framework to recover location labels directly from sequences of sparse channel state information (CSI) measurements, without explicit location labels for radio map construction. Instead of directly storing raw CSI, we learn a compact low-dimensional radio map embedding and leverage a generative model to reconstruct the high-dimensional CSI. Specifically, to address the uncertainty of sparse CSI, a dual-scale feature extraction scheme is designed to enhance feature representation by jointly exploiting correlations from angular space and across neighboring samples. We develop a hybrid recurrent-convolutional encoder to learn mobility patterns, which combines a truncation strategy and multi-scale convolutions in the recurrent neural network (RNN) to ensure feature robustness against short-term fluctuations. Unlike conventional Gaussian priors in latent space, we embed a learnable radio map to capture the location information by encoding high-level positional features from CSI measurements. Finally, a diffusion-based generative decoder reconstructs the full CSI with high fidelity by conditioning on the positional features in the radio map. Numerical experiments demonstrate that the proposed model can improve localization accuracy by over 30% and achieve a 20% capacity gain in non-line-of-sight (NLOS) scenarios compared with model-based Kalman filter approaches.

Figures (12)

• Figure 1: Diagram of a radio-map-embedded generative architecture.
• Figure 2: The proposed hybrid RNN-CNN based CSI encoder for joint trajectory recovery and radio map learning.
• Figure 3: The proposed generative decoder conditioned on the radio map for full CSI reconstruction.
• Figure 4: Environmental topology: a) Scenario I: 3D terrain of the outdoor urban environment; b) Scenario II: Top view of the indoor environment dataset-dichasus-cf0x.
• Figure 5: Trajectory recovery results for the outdoor scenario, where the moves continuously at 1 m/s along the main road in a back-and-forth pattern.