Load Balancing in Small-Cell Access Point Placement

We address the uplink small-cell access point (AP) placement problem for optimal throughput, while considering load balancing (LB) among the APs. To consider LB and consequently incorporate fairness in user spectral access, i.e., the frequency of user-to-AP communications, we modify the Lloyd algorithm from vector quantization so that delays incurred by the existence of a large number of users in a cell are accounted for in the AP placement process. Accordingly, we present two methods, the first of which involves the incorporation of weights proportional to the cell occupancy, hence called the Occupancy Weighted Lloyd algorithm (OWLA). The second method adds a new step to the Lloyd algorithm, which involves re-assigning users from higher to lower occupancy cells, and the adoption of a distance threshold to cap the throughput lost in the assignment process. This formulated Lloyd-type algorithm is called the Cell Equalized Lloyd Algorithm-α (CELA-α) where α is a factor that allows for throughput and spectrum access delay trade-off. Extensive simulations show that both CELA-α and OWLA algorithms provide significant gains, in comparison to the standard Lloyd algorithm, in 95%-likely user spectral access. For the α values considered in this paper, CELA-α achieves gains up to 20.83%, while OWLA yields a gain of 12.5%. Both algorithms incur minimal throughput losses of different degrees, and the choice of using one algorithm over the other for AP placement depends on system LB as well as throughput requirements.