Self-Supervised Learning for User Localization

TL;DR

This work tackles 3D user localization from Channel State Information (CSI) under scarce labeled data. It introduces self-supervised pretraining using autoencoders (MLP-based and CNN-based) on unlabeled CSI to learn robust representations, which are then fed into a downstream MLP for 3D position estimation of users. On the CTW-2020 dataset, the CNN-based pretraining approach achieves the best performance with an average MAE of approximately $16.87$ meters, outperforming purely supervised baselines by a substantial margin. The study demonstrates that exploiting unlabeled CSI through self-supervised learning improves robustness and generalization for large-area localization tasks, particularly when labeled data are limited.

Abstract

Machine learning techniques have shown remarkable accuracy in localization tasks, but their dependency on vast amounts of labeled data, particularly Channel State Information (CSI) and corresponding coordinates, remains a bottleneck. Self-supervised learning techniques alleviate the need for labeled data, a potential that remains largely untapped and underexplored in existing research. Addressing this gap, we propose a pioneering approach that leverages self-supervised pretraining on unlabeled data to boost the performance of supervised learning for user localization based on CSI. We introduce two pretraining Auto Encoder (AE) models employing Multi Layer Perceptrons (MLPs) and Convolutional Neural Networks (CNNs) to glean representations from unlabeled data via self-supervised learning. Following this, we utilize the encoder portion of the AE models to extract relevant features from labeled data, and finetune an MLP-based Position Estimation Model to accurately deduce user locations. Our experimentation on the CTW-2020 dataset, which features a substantial volume of unlabeled data but limited labeled samples, demonstrates the viability of our approach. Notably, the dataset covers a vast area spanning over 646x943x41 meters, and our approach demonstrates promising results even for such expansive localization tasks.

Figures (3)

• Figure 1: User Equipment position on the XY plane with dimensions in meters and base station at (0,0).
• Figure 2: Model 1 and Model 2 architecture for Supervised Learning with labeled data.
• Figure 3: Model 3 and Model 4 architecture for pretraining and finetuning with unlabeled and labeled data.