Learning-based efficient access methods for machine-type communications: Multi-user detection
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
Saalasti Hall, Linnanmaa
Topic of the dissertation
Learning-based efficient access methods for machine-type communications: Multi-user detection
Doctoral candidate
Master of Science (Technology) Thushan Sivalingam
Faculty and unit
University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, CWC-Radio Technologies
Subject of study
Communications Engineering
Opponent
Professor Arumugam Nallanathan, Queen Mary University, London
Custos
Professor Nandana Rajatheva, University of Oulu
Learning-based efficient access methods for machine-type communications: Multi-user detection
The objective of this thesis is to present efficient, novel learning-based access methods for massive machine-type communications (mMTC). mMTC is envisioned as one of the leading service classes for sixth-generation (6G) wireless communication systems. It will focus on supporting massive numbers of uplink-dominated, low-power, and low-complexity devices that sporadically transmit short data packets at low transmission rates during short active state periods, posing challenges in efficient random access. Grant-free random access and uplink non-orthogonal multiple access (NOMA) are introduced to increase the overload factor and reduce transmission latency while minimizing signaling overhead. Sparse code multiple access (SCMA) and Multi-user shared access (MUSA) are introduced as advanced code domain NOMA schemes. In grant-free NOMA, machine-type devices (MTD) transmit information to the base station (BS) without a grant, creating a challenging task for the BS to identify the active MTDs among all potential active devices, referred to as multi-user detection (MUD).
A novel pre-activated residual neural network-based MUD scheme for the grant-free SCMA and MUSA system in an mMTC uplink framework is proposed to jointly identify the number of active MTDs in the absence of channel state information (CSI) and device sparsity. A novel residual unit designed to learn the properties of multi-dimensional SCMA codebooks, MUSA spreading sequences, and corresponding combinations of active devices with single measurement vector (SMV) and multiple measurement vector (MMV) systems. Furthermore, an MUD approach applied to spatially correlated Rician fading channels is formulated and subsequently reformulated as a multi-label classification problem utilizing deep learning techniques. Two diverse approaches have been proposed to tackle this problem: ViT-Net, a vision transformer-based architecture, and FAR-Net, a fully activated deep neural network featuring residual connections.
A realistic indoor manufacturing environment, including indoor mobility scenarios with experimental data-based MUD scenarios, is considered for investigating low-complexity algorithms. Empirical evidence supports the approach by conducting a channel measurement campaign in a functional indoor factory with standard parameters and universal software radio peripheral (USRP) modules. Numerical evaluations demonstrate the effectiveness of the proposed schemes compared to existing approaches over the signal-to-noise ratio range of interest.
A novel pre-activated residual neural network-based MUD scheme for the grant-free SCMA and MUSA system in an mMTC uplink framework is proposed to jointly identify the number of active MTDs in the absence of channel state information (CSI) and device sparsity. A novel residual unit designed to learn the properties of multi-dimensional SCMA codebooks, MUSA spreading sequences, and corresponding combinations of active devices with single measurement vector (SMV) and multiple measurement vector (MMV) systems. Furthermore, an MUD approach applied to spatially correlated Rician fading channels is formulated and subsequently reformulated as a multi-label classification problem utilizing deep learning techniques. Two diverse approaches have been proposed to tackle this problem: ViT-Net, a vision transformer-based architecture, and FAR-Net, a fully activated deep neural network featuring residual connections.
A realistic indoor manufacturing environment, including indoor mobility scenarios with experimental data-based MUD scenarios, is considered for investigating low-complexity algorithms. Empirical evidence supports the approach by conducting a channel measurement campaign in a functional indoor factory with standard parameters and universal software radio peripheral (USRP) modules. Numerical evaluations demonstrate the effectiveness of the proposed schemes compared to existing approaches over the signal-to-noise ratio range of interest.
Last updated: 19.8.2025