DeepMuD: Multi-user Detection for Uplink Grant-Free NOMA IoT Networks via Deep Learning

TL;DR

This paper introduces DeepMuD, a deep learning-based multi-user detection method for uplink NOMA IoT networks that does not require perfect CSI and can handle an arbitrary number of devices, significantly improving error performance.

Contribution

The paper presents a novel LSTM-based deep learning approach for joint channel estimation and multi-user detection in uplink NOMA, enabling grant-free IoT communication without signaling overhead.

Findings

DeepMuD outperforms conventional detectors even with imperfect CSI.

The method scales well with increasing number of IoT devices.

It enables grant-free communication with flexible detection regardless of device number.

Abstract

In this letter, we propose a deep learning-aided multi-user detection (DeepMuD) in uplink non-orthogonal multiple access (NOMA) to empower the massive machine-type communication where an offline-trained Long Short-Term Memory (LSTM)-based network is used for multi-user detection. In the proposed DeepMuD, a perfect channel state information (CSI) is also not required since it is able to perform a joint channel estimation and multi-user detection with the pilot responses, where the pilot-to-frame ratio is very low. The proposed DeepMuD improves the error performance of the uplink NOMA significantly and outperforms the conventional detectors (even with perfect CSI). Moreover, this gain becomes superb with the increase in the number of Internet of Things (IoT) devices. Furthermore, the proposed DeepMuD has a flexible detection and regardless of the number of IoT devices, the multi-user detection can be performed. Thus, an arbitrary number of IoT devices can be served without a signaling overhead, which enables the grant-free communication.