Resource management for cellular networks in shared spectrum

We are witnessing tremendous growth in mobile data traffic. Ericsson predicts [34] that the total mobile traffic is expected to increase by a factor of 5 over the next 6 years, reaching 136 EB/month by the end of 2024. Given this high capacity requirement and constraints on spectrum access because of traditional regulatory approaches, cellular network providers have started to feel the spectrum crunch. In order to address the scarcity of spectrum, we explore the use of cellular radio technology in the shared spectrum. There are several challenges to realize this. First of all, cellular technology, such as LTE, has been designed for planned operator deployments. They dont support unplanned interference management, interoperability with spectrum databases, etc. Secondly, assigning optimal channel resources to interfering nodes requires precise channel estimation and prediction which is very challenging in current LTE systems especially when clients are highly mobile. In this dissertation we design techniques to support unplanned interference management and robust channel estimation and prediction scheme for LTE. First, we study an LTE based network architecture for unlicensed cellular networking in shared spectrum. We design an LTE-compatible interference management component for distributed coordination among users operating in the same frequency channel. Our proposed interference management requires no explicit communication between base stations, and is more efficient than CSMA for long links. Second, we develop a decentralized spectrum management system for unlicensed LTE users in a shared band. We design a channel allocation scheme to efficiently utilize spectrum and incentivise collaboration. This also includes a simple, non-disruptive channel change scheme to frequently and efficiently change channels to accommodate dynamic traffic and environments. Third, we develop a novel channel estimation and prediction algorithm for next generation cellular networks. Instead of directly predicting the channel in the widely used frequency time domain, we predict the channel in the delay-Doppler domain. Our intuition is that mobility is easier to predict than the channel since mobility has inertial correlation, whereas wireless channel is the outcome of complicated interaction between mobility, multipath, and noise. Reliably estimating and predicting the wireless channel is necessary for optimal resource allocation among highly mobile nodes operating in shared spectrumComputer Science