María de los Ángeles Simarro Haro

This thesis fits into the Multiple-Input Multiple-Output (MIMO) communication systems. Nowadays, these schemes are the most promising technology in the field of wireless communications, since they let to take advantage of the spatial dimension together with the dimensions of frequency and time. In this way, the use of this technology allows to increase the rate and the quality of the transmission through the use of multiple antennas at the transmitter and receiver sides. Furthermore, the MIMO technology can also be used in a multiuser scenario, where a Base Station (BS) equipped with several antennas serves several users that share the spatial dimension causing interference. However, employing precoding algorithms the signal of the multiuser interference can be mitigated. For these reasons, the MIMO technology has become an essential key in many new generation communications standards such as Wireless Local Area Network (WLAN), Worldwide interoperability for Microwave Acces (WiMAX), Long Term Evolution (LTE) or Next Generation Handheld (DVB-NGH). On the other hand, Massive MIMO technology or Large MIMO, where the BS is equipped with very large number of antennas (hundreds or thousands) serves many users in the same time-frequency resource, it is a promising candidate technology for next generations of wireless systems. Nevertheless, the advantages provided by the MIMO technology entail a substantial increase in the computational cost. Therefore the design of low-complexity receivers is an important issue which is tackled throughout this thesis. To this end, one of the main contributions of this dissertation is the implementation of efficient soft-output detectors, since this stage is considered the most complex part of the communication process.

On the other hand, in a multiuser scenario the amount of computational cost is carried out by the precoding processing in the BS, allowing the development of small and inexpensive terminals. Therefore, other important contribution of this thesis is focused on improving the efficiency of precoding schemes. First, the problem of efficient soft detection with no iteration at the receiver has been addressed. That is, detectors which only process the received vector. A detailed overview of the most employed soft detectors is provided. Furthermore, the complexity and performance of these methodsare evaluated and compared. Additionally, two low-complexity algorithms have been proposed. The first algorithm is based on the efficient Box Optimization Hard Detector (BOHD) algorithm and provides a low-complexity implementation achieving a suitable performance. The second algorithm tries to reduce the computational cost of the Subspace Marginalizationwith Interference Suppression (SUMIS) algorithm. Second, soft-input soft-output (SISO) detectors, which are included in an iterative receiver structure, have been investigated. A SISO detector processes the information received by the channel and also the information provided by the channel decoder in the feedback loop. An iterative receiver improves the performance with respect to no iteration, achieving a performance close to the channel capacity. In contrast, its computational cost becomes prohibitive. In this context, three algorithms are presented. Two of them achieve max-log performance reducing the complexity of standard SISO detectors. The last one achieves near max-log performance with low complexity. The precoding problem has been addressed in the third part of this thesis. An analysis of some of the most employed precoding techniques has been carried out. The algorithms have been compared in terms of performance and complexity. In this context, the impact of the channel matrix condition number on the performance of the precoders has been analyzed. This impact has been exploited to propose an hybrid precoding scheme that reduces the complexity of the previously proposed precoders. In addition, in Large MIMO systems, an alternative precoder scheme is proposed. The proposed scheme reduces the computational cost with respect to the conventional precoding algorithms while good performance is maintained.

In the last part of the thesis, parallel implementations of the SUMIS algorithm are presented. Several strategies for the parallelization of the algorithm are proposed and evaluated on two different platforms: multicore central processing unit (CPU) and graphics processing unit (GPU). The parallel implementations achieve a significant speedup compared to the CPU version when the number of antennas or constellation order increase, that is the context of Large MIMO. Therefore, these implementations allow to simulate a scalable quasi optimal soft detector in a Large MIMO system much faster than by conventional simulation.