Optimum Detection Techniques for MIMO and Massive MIMO Systems

Abstract

High throughput, energy efficiency, and connectivity of a massive number of users are the main requirements of existing wireless communication systems. The three basic and commonly used techniques that can fulfil the spectral and energy efficiency requirements of these networks include finding new frequency resources, deploying ultra-dense networks, and using multiple antenna systems. In recent years, Multiple Input Multiple Output (MIMO) systems have appeared as a promising, reliable, and spatially effectual solution to the growing shortage of radio frequency communications bands. These systems, when employed for spatial multiplexing, provide a rise in capacity without the requirement for extra spectrum and power. Their improved data rates and performance have made them increasingly prominent in modern wireless devices. But with the decreased bit error rate, increased diversity, and array gains of MIMO systems, comes an increased complexity in the receiver design. An optimal design of transmitter and receiver is required to enjoy the advantages presented by these systems. For fifth generation and beyond communication systems, massive MIMO systems are considered a cutting edge technology. It is a promising technique for providing orders of magnitude improvements in throughput, coverage, energy, and spectral efficiency. Having several antennas at the base station, the massive MIMO system achieves prominent performance advantages. The problem with massive MIMO systems is that the signal processing complexity increases exponentially with a large system limit. It becomes challenging to extract the individual signals from the composite signal, thus making the optimal detectors prohibitively complex. For both MIMO and massive MIMO systems, linear and non-linear detectors are proposed in the literature for symbol detection. Linear detectors like Zero Forcing and Minimum Mean Squared Error (MMSE) are computationally less complex than non-linear detectors like Vertical Bell-Labs Layered Space-Time. However, they suffer from high bit error rate performance. The drawback of linear detectors is that they perform matrix inversion operations, which are not hardware friendly. To address this issue and reduce the computational complexity of linear detectors even further, approximate detectors, e.g., Gauss-Seidel, Neumann Series, etc., are proposed in the literature. However, they show less satisfactory performances. The effectiveness of massive MIMO detectors relies mostly on the channel state information. Therefore, the base station has to be aware of it to ensure good quality of communication. The majority of research done in the literature considers perfect channel state information at the receiver for evaluation of signal to noise and interference ratio, symbol error rate, and bit error rate performances. Practically, it is hard to attain perfect channel state information because of feedback delays and imperfections in information extraction, resulting in imperfect channel state information. Also, an infeasible number of pilot signals, reciprocity errors, and electromagnetic coupling between a large number of antennas at base stations with inadequate spacing impact the overall efficiency. There is a paucity of work in the literature covering analysis of massive MIMO detectors under imperfect channel state information. Little information is present on the bit error rate analysis of detectors under imperfect channel state information. Firstly, in this thesis, detectors for MIMO systems are investigated, and an algorithm based on local search is presented to improve the performance of existing detectors. The proposed algorithm yields improved bit error rate performance of linear detectors. Secondly, a hybrid detector consisting of linear and approximate detector is proposed for massive MIMO systems. The proposed detector, a hybrid Neumann Series based MMSE assisted detector, is simulated in a Rayleigh fading channel and evaluated along with approximate message passing algorithm having Ternary and Gaussian distribution threshold functions. Simulation results confirm that the proposed hybrid Neumann Series based MMSE assisted detection algorithm performs significantly better than the mentioned detection schemes. Finally, the performance of approximate detectors, i.e., Gauss-Seidel and Neumann Series, is evaluated under imperfect channel state information. The results are compared with the linear detectors under imperfect channel state information. It is noted with the help of Monte-Carlo numerical simulations, that under imperfect channel state information, the Gauss-Seidel detector gives comparative bit error rate performance as the linear detectors with lower computational complexity.