| Summary: | <p>Mobile data traffic has experienced exponential growth in recent decades, which imposes increasing pressure on the limited resources of wireless communications. Meanwhile, the increased use of white light-emitting diodes (LEDs) has led to a growing interest in a novel wireless communications technology: Indoor visible light communication (VLC). The unlicensed spectrum provided by VLC supplies the additional capacity to mobile communication systems, while the LEDs can also offer illumination. This illumination is often provided by multiple LEDs installed on the ceiling in an indoor environment, which offers the potential to develop Multiple-Input Multiple-Output (MIMO) VLC systems. MIMO systems are subject to crosstalk between the multiple communications channels and mitigating this crosstalk is an active area of research. This thesis focuses on how to best use the MIMO resource regarding the complexity of implementation and control of such a system.</p>
<p>First, the work focuses on a high-speed indoor MIMO-VLC system that optimises the number of MIMO channels and the data rates. A Channel Condition-based Transmitter Coordination (CCTC) algorithm is proposed. In the CCTC algorithm, transmitters are coordinated to form one or several group(s) based on the wireless channel conditions, then the data rate of each group is optimised separately. The system employs zero-forcing MIMO equalisation and decision-feedback equalisation (DFE) to mitigate crosstalk. Two receiver designs have also been employed: A seven-channel angle diversity receiver (ADR) and an aperture-based receiver (ABR). In the new system, the overall data rates are on average 41 per cent (with an ADR) and 127 per cent (with an ABR) higher than that in the conventional spatial multiplexing approach. Moreover, the peak overall data rate of the system with an ADR (676 Mbps) is 56 per cent higher than that with an ABR (434 Mbps), which means that an ADR shows better performance than an ABR when they are in a similar size. The proposed approach efficiently mitigates the crosstalk and improves the data rates in the MIMO system. However, it employs exhaustive search to coordinate transmitters and data rates, which takes a large amount of computing time.</p>
<p>A solution with less computational complexity is therefore explored. A neural network is employed to replace the proposed algorithm to obtain optimal results. The neural network is trained and tested with varying receiver positions, yaw rotation angles, or roll rotation angles relative to the lighting sources. A statistical model with random orientations and receiver positions depicting a user walking in the coverage area is also employed. The results show that the neural network-based system saves a factor of three to four in computing time compared with the CCTC algorithm system while achieving the same level of performance. The water-filling algorithm is employed as a benchmark of the maximum capacity of the system to assess the neural network performance. The neural network only takes one-third of the computing time compared to the water-filling algorithm, with a 12 per cent data rate penalty on average. Moreover, it is simpler in terms of control and implementation than the water-filling algorithm. These results show the potential of the more straightforward neural network method, and future directions for this work are presented in the conclusions of the thesis.</p>
|
|---|