Three-Dimensional Radio Localization: A Channel Charting-Based Approach

TL;DR

This work tackles 3D indoor localization by leveraging channel charting to produce self-supervised, geometry-preserving representations of the radio environment. It introduces augmented channel charting to ground the chart in physical space and a multistory extension that uses floor clustering plus per-floor experts, along with novel beamspace CSI features to boost localization. Through ray-tracing datasets, conventional channel charting (with alignment) achieves sub-meter MAE in 3D scenarios, while multistory channel charting further improves accuracy to ~1 m with a small floor-classification error. The methods show promise for robust 3D localization in indoor settings, with limitations noted for ground-truth validation and density requirements, suggesting directions for real-world validation and broader environmental testing.

Abstract

Channel charting creates a low-dimensional representation of the radio environment in a self-supervised manner using manifold learning. Preserving relative spatial distances in the latent space, channel charting is well suited to support user localization. While prior work on channel charting has mainly focused on two-dimensional scenarios, real-world environments are inherently three-dimensional. In this work, we investigate two distinct three-dimensional indoor localization scenarios using simulated, but realistic ray tracing-based datasets: a factory hall with a three-dimensional spatial distribution of datapoints, and a multistory building where each floor exhibits a two-dimensional datapoint distribution. For the first scenario, we apply the concept of augmented channel charting, which combines classical localization and channel charting, to a three-dimensional setting. For the second scenario, we introduce multistory channel charting, a two-stage approach consisting of floor classification via clustering followed by the training of a dedicated expert neural network for channel charting on each individual floor, thereby enhancing the channel charting performance. In addition, we propose a novel feature engineering method designed to extract sparse features from the beamspace channel state information that are suitable for localization.

Figures (7)

• Figure 1: Visualization of the ray tracing-based datasets: The figure shows (a) an image of the synthetic factory hall (scenario 1), and (c) an image of the synthetic multistory building (scenario 2), both with the BS antenna arrays marked as green dots and the "ground truth" UE positions as red dots, and (b) and (d) the respective scatter plots of the "ground truth" positions, each with a unique color gradient. (b) additionally marks the array positions as black dots.
• Figure 2: Conventional and multistory channel charting pipeline.
• Figure 3: Neural network structures: (a) DNN as forward charting function, and (b) Siamese network for the training process of channel charting.
• Figure 4: Scatter plots of the position estimates for scenario 1 with color gradient preserved from the ground truth positions in Fig. \ref{['fig:dataset_scenario1_positions']}: (a) classical localization, (b) channel charting, (c) channel charting followed by an optimal affine tranformation w.r.t. the ground truth positions, and (d) augmented channel charting.
• Figure 5: Scenario 1: MAE vs. elevation resolution.