UC San Diego

The past decade has witnessed the surge of wireless communication technologies, one of which is massive multiple-input-multiple-output (MIMO) which forms a prominent part of 5th generation (5G) and future generations of wireless networks. Among distributed massive MIMO networks, where there are a very large number of antennas physically spread out across a geographical area, one of the degrees of freedom available is access point (AP) placement. In both the traditional small-cell and the more recent cell-free networks, AP placement can be a key enabler of higher throughput or spectral efficiency. Thus, in this dissertation, we study AP placement in both small-cell and cell-free regimes by examining the throughput maximization problem. With the vector quantization (VQ) approach showing similarities to AP placement, we study and utilize it as a starting point for our framework on AP placement formulations.

First, the AP placement problem in the small-cell scenario that addresses signal-to-interference-plus-noise ratio (SINR) maximization (as opposed to signal-to-noise ratio (SNR) maximization alone) is studied since inter-cell interference (ICI) is a prevalent feature of small-cell systems. By first establishing that the Lloyd algorithm from VQ that utilizes the squared Euclidean distance as the distortion measure forms a good enough solution for AP placement, the distortion is modified to include inter-cell interference (ICI) and a Lloyd-type algorithm is proposed to solve for the AP locations.

Second, the placement solution is extended to hybrid networks consisting of both terrestrial (position-fixed) and unmanned aerial vehicle (UAV) enabled (position-flexible) APs. In this scenario, a Lloyd-type algorithm is used to solve for the positions of the UAV-APs while still accounting for ICI. Additionally, a solution to initialize AP locations for the Lloyd and Lloyd-type algorithms is also discussed. As a further extension, we consider load balancing, which is necessary due to the varied user-AP access resulting from unequal cell occupancies. To address this, two Lloyd-type algorithms are proposed to make the cell occupancies more equal and hence enabling fairness in user-AP access.

Third, the throughput problems for AP placement in cell-free networks are studied, namely the sum rate and minimum rate maximization problems. Their solution structures are analyzed and simple supporting examples are provided. The VQ framework is motivated as a practical approach for these problems. Three VQ-based techniques, namely the Lloyd algorithm, the tree-structured VQ (TSVQ), and probability density function optimized VQ (PDFVQ), each with its own advantages for cell-free AP placement, are outlined and their performances compared.

Finally, in addition to the AP placement and to make the cell-free system more viable, two important and necessary features, namely user position determination and limited cooperation, are added. A multi-step AP deployment process is designed and demonstrated incorporating these features.