Low-Complexity Near-Optimal Detection Algorithms for MIMO Systems
Abstract
As the number of subscribers in wireless networks and their demanding data rate are exponentially increasing, multiple-input multiple-output (MIMO) systems have been scaled up in the 5G where tens to hundreds of antennas are deployed at base stations (BSs). However, by scaling up the MIMO systems, designing detectors with low computational complexity and close to the optimal error performance becomes challenging. In this dissertation, we study the problem of efficient detector designs for MIMO systems.
In Chapter 2, we propose efficient detection algorithms for small and moderate MIMO systems by using lattice reduction and subspace (or conditional) detection techniques. The proposed algorithms exhibit full receive diversity and approach the bit error rate (BER) of the optimal maximum likelihood (ML) solution. For quasi-static channels, the complexity of the proposed schemes is cubic in the system dimension and is only linear in the size of the QAM modulation used. However, the computational complexity of lattice reduction algorithms imposes a large burden on the proposed detectors for large MIMO systems or fast fading channels.
In Chapter 3, we propose detectors for large MIMO systems based on the combination of minimum mean square error decision feedback equalization (MMSE-DFE) and subspace detection tailored to an appropriate channel ordering. Although the achieved diversity order of the proposed detectors does not necessarily equal the full receive diversity for some MIMO systems, the coding gain allows for close to ML error performance at practical values of signal-to-noise ratio (SNR) at the cost of a small computational complexity increase over the classical MMSE- DFE detection. The receive diversity deficiency is addressed by proposing another algorithm in which a partial lattice reduction (PLR) technique is deployed to improve the diversity order.
Massive multiuser MIMO (MU-MIMO) is another technology where the BS is equipped with hundreds of antennas and serves tens of single-antenna user terminals (UTs). For the uplink of massive MIMO systems, linear detectors, such as zero-forcing (ZF) and minimum mean square error (MMSE), approach the error performances of sophisticated nonlinear detectors. However, the exact solutions of ZF and MMSE involve matrix-matrix multiplication and matrix inversion operations which are expensive for massive MIMO systems. In Chapter 4, we propose efficient truncated polynomial expansion (TPE)-based detectors that achieve the error performance of the exact solutions with a computational complexity proportional to the system dimensions.
The millimeter wave (mmWave) massive MIMO is another key technology for 5G cellular networks. By using hybrid beamforming techniques in which a few numbers of radio frequency (RF) chains are deployed at the BSs and the UTs, the fully-digital precoder (combiner) is approximated as a product of analog and digital precoders (combiners). In Chapter 5, we consider a signal detection scheme using the equivalent channel consisting of the precoder, mmWave channel, and combiner. The available structure in the equivalent channel enables us to achieve the BER of the optimal ML solution with a significant reduction in the computational complexity.
Collections
Cite this version of the work
Kazem Izadinasab
(2020).
Low-Complexity Near-Optimal Detection Algorithms for MIMO Systems. UWSpace.
http://hdl.handle.net/10012/16401
Other formats
Related items
Showing items related by title, author, creator and subject.
-
Conflict detection in dual-process theory: Are we good at detecting when we are biased at decision making?
Pennycook, Gordon Robert (University of Waterloo, 2011-08-31)In the domain of reasoning and decision making, some dual-process theorists have suggested that people are highly efficient at detecting conflicting outputs engendered by competing intuitive and analytic processes (De Neys ... -
Out-of-Distribution Detection for LiDAR-based 3D Object Detection
Nguyen, Van Duong (University of Waterloo, 2022-01-18)3D object detection is an essential part of automated driving, and deep neural networks (DNNs) have achieved state-of-the-art performance for this task. However, deep models are notorious for assigning high confidence ... -
Predictive models and detection methods applicable in water detection framework for industrial electric arc furnaces
Alshawarghi, Hamzah; Elkamel, Ali; Moshiri, Behzad; Hourfar, Farzad (Elsevier, 2019-09-02)This paper introduces the development of empirical predictive models and detection methods that are incorporated into a water detection framework for an industrial steelmaking electric arc furnace (EAF). The predictive ...