Distributed Interference Management in Unplanned Wireless Environments

Abstract

As new applications for wireless communications are developed, system designers are turning more frequently to unplanned networks. While cheaper to deploy, such
networks suffer a deficit in control and planning which have a negative impact on the network's ability to manage interference. The lack of centralised power control and scheduling increases the impact of collisions which manifest as in-band interference, negatively impacting performance. This thesis examines several aspects of interference management and determines the parameters that a system designer can manipulate to control the behaviour of the network. First we determine the optimal update rate for a subspace-based distributed interference alignment scheme, reducing convergence time to within 100 iterations. While this improves the convergence time of the algorithm it offers no capacity improvement due to a lack of power-awareness in the algorithm. An altered algorithm is then proposed that offers power-awareness and is found to improve convergence time to less than 10 iterations at a cost to sum interference levels and network capacity. We then propose training sequences that enables users to determine the number of antennas another user possesses with less than 1% error, using this information to recover the training signal for channel estimation. These sequences are found to
perform better than the equivalent chaotic training sequences in estimating channel coeffcients. This approach continues to function in channels with unequal transmit and receive antennas, permitting its use in heterogeneous networks. Finally we derive an analytic expression for the upper bound of the likelihood an arbitrary number of neighbours can be detected by a user using SIC on transmissions from directional antenna arrays during neighbour discovery. The expression found permits the system designer to optimise system parameters such as SIR threshold, user array directivity, and transmission rate given the expected user density of the network.

Date of Award	26 Nov 2020
Original language	English
Awarding Institution	The University of Bristol
Supervisor	Simon M D Armour (Supervisor), Angela Doufexi (Supervisor), Woon Hau Chin (Supervisor) & Filippo Tosato (Supervisor)