Obstacle Detection and Infrastructure Deployment for Efficient Millimeter-wave Communications

Abstract

There has been growing interest in millimeter-wave (mmWave) communication due to the promising high speeds and immense amounts of unused bandwidth available. However, mmWaves suffer from unusually high attenuation, through free space, and especially through obstacles, which necessitates an obstacle free line-of-sight (LOS) transmission path. This thesis deals with establishment of such LOS paths, through obstacle detection and deployment of network infrastructure. The usual approach to avoid static obstacles on transmission paths is to use satellite imagery to detect the presence of static obstacles, an approach which apart from raising proprietary concerns, is not able to capture smaller obstacles. We propose a simple learning based approach to detect the presence of static as well as dynamic obstacles, without having apriori access to any data regarding their location from satellite imagery. We then use this knowledge to efficiently select an appropriate transmission path for a user equipment (UE), lowering the chance of allocating an obstacle prone link. Dynamic obstacles are usually tracked by dedicated tracking hardware like RGB-D cameras, which usually have small ranges, and hence lead to prohibitively increased deployment costs to achieve complete camera coverage of the deployment area. We propose an altogether different approach to track dynamic obstacles in an mmWave network, solely based on short-term historical link failure information, without resorting to any dedicated tracking hardware. Using the obtained trajectories, we perform proactive handoffs for at-risk links. We compare our approach with an RGB-D camera-based approach and show that our approach provides better handoff performances when the camera coverage is low to moderate, which is often the case in real deployment scenarios. Stability of allocated transmission paths is an important problem in the domain of mmWave communication. The quality of an allocated transmission path depends not only upon the present time, but also upon the maintenance of the said path in the near future; the fragile nature of mmWaves necessitates this. Thus, allocating the base station (BS) which provides the highest received signal strength (RSS) at the current time instant is not always the best idea, considering UE mobility, and presence of obstacles. We propose a simple geometric approach to allocate stable transmission paths which are less likely to be broken in the near future. One way to deal with obstacle free strict LOS requirements of mmWaves is to densely deploy small range mmWave BSs, to overcome outage due to obstacles. Low cost reflectors have also been proposed to augment the transmission environment, and reflect mmWaves in the desired direction, thereby bypassing the obstacles. We argue that considering the placement of mmWave BSs and reflectors independently may lead to suboptimal coverage. We consider an urban deployment scenario, and attempt to maximally cover it by jointly placing the mmWave BSs and reflectors. Given the hardness of the joint problem, we first develop a set cover based greedy solution, and also provide a linear programming (LP) relaxation based solution. With extensive simulations, we show that with a fixed number of available mmWave BSs and reflectors to be placed, both our proposed solutions achieve a larger coverage compared to an existing approach where BSs and reflectors were placed sequentially. Unmanned Aerial Vehicles (UAVs) are a potential platform for deployingmmWave BSs. One challenge that has to be addressed is the limited power onboard a UAV, which is used to hover and move the device, and of course, to transmit data. We deal with the deployment of UAVs with an aim to minimise their displacement in subsequent time instances. We take into consideration UE mobility, and propose LazyUAV, a set cover based geometric approach to minimise UAV displacement, while maintaining maximal coverage.