Text

thesis

ETD doctoral

The high utilization of the currently licensed spectrum for mobile communications, especially in dense urban areas, has led spectrum regulators to explore new ways to alleviate this congestion issue. One proposed solution is to exploit higher frequency bands, e.g., the millimeter wave (mmWave) spectrum, in future generations of wireless networks. Although mmWave signals can increase the network capacity and achieve high data rates, they are sensitive to blockage and attenuation, and thus their deployment in real settings presents challenges. Relaying in combination with beamforming is a promising way to mitigate those effects and increase communication range. As the wireless channel changes frequently in both space and time, a relay-beamforming network must constantly adapt to those changes in order to guarantee a desired Quality-of-Service (QoS) to its network users. In this work we consider two different scenarios, i.e., static relays and mobile relays. Determining the optimal configuration among static relays, or controlling the movement of mobile relays in order to meet a certain QoS, requires knowledge of time and location specific Channel State Information (CSI), an at best resource demanding task that also introduces additional network latency. The subject of this dissertation is focused on developing predictive schemes to determine the optimal relay configuration.

An innovative cooperative relay beamforming approach is proposed to support mmWave wireless communications in cities, that relies on an infrastructure of static relay clusters deployed within the urban environment. In particular, assuming a time-slotted system operation, one relay from each cluster is optimally selected at each time slot to participate in optimal beamforming at the next time slot, in order to maximize the expected Signal-to-Interference plus Noise Ratio (SINR) at the destination, under power constraints. The key novelty of the proposed scheme is that relay selection is implemented in a predictive and distributed manner, i.e., each cluster independently decides its beamforming relay for the subsequent time slot without the need for inter-cluster information exchange, by exploiting channel correlations and by using past and present measurements of magnitude CSI. The proposed relay selection scheme is executed in parallel to the beamforming, eliminating delays induced by the otherwise sequential execution of relay selection and beamforming, and substantially reducing the required CSI estimation overhead, as compared to conventional methods.

We also examine the case where a set of mobile relays support the communication of a single or multiple source-destination pairs. In this case determining the optimal relay configuration is essentially a relay motion control problem, since the mobile relays attempt to learn a sequence of actions, namely a policy, that will allow them to be optimally positioned to beamform. A time slotted approach is again considered, where in each slot the relays implement optimal beamforming in order to achieve the desired QoS, e.g., maximizing the weakest SINR at the destinations or minimizing the total relay transmit power, and in parallel they must estimate in a predictive fashion their optimal positions for the next slot. Our proposed motion control policies are based on two different learning paradigms, model-based and model-free learning. In the model-based policy, the underlying channel model is assumed to be known to the relays. Based on that model, the relays estimate key channel state parameters in order to predict the magnitude CSI of all candidate future locations and, based on those predictions, execute their movement decisions. In the second model-free policy, prediction of channel state parameters is not required as the relay movement decisions are guided using Q-learning, i.e., according to the quality values of each possible relay state-movement combination.

ETD_11365

Ph.D.

Includes bibliographical references

doi:10.7282/t3-6np1-ds41

The author owns the copyright to this work.