Year: 2020
Type: dissertation
Abstract: This thesis considers the massive random access problem in which a large number of sporadically active devices attempt to communicate with the base-station (BS) in the uplink. For device identification, each device is preassigned a unique signature sequence from a large set of non-orthogonal sequences as its identifier. In random access, all active devices transmit their sequences simultaneously, and the BS identifies the active devices by detecting which sequences are transmitted. This thesis focuses on the device activity detection, and studies two fundamentally distinct problem formulations: the compressed sensing approach, which aims to attain the activity by recovering the effective instantaneous channel; and the covariance approach, which aims to estimate the effective channel statistics for activity detection. This thesis shows the covariance approach is better suited for massive multiple-input multiple-output (MIMO), whereas the compressed sensing approach is favored with long sequences. For the compressed sensing approach, this thesis considers single-cell and multi-cell scenarios. In both cases, the device activity detection problem is formulated as a compressed sensing problem, and the computationally efficient approximate message passing (AMP) algorithm is employed. Specifically, for the single-cell scenario, this thesis studies the design of the AMP based detector to incorporate prior information of wireless channel, and also provides an analytical characterization of the probabilities of false alarm and missed detection. For the multi-cell scenario, this thesis compares two network architectures, cooperative MIMO and massive MIMO, in terms of their effectiveness in combating inter-cell interference for device activity detection. For the covariance approach, this thesis considers a single-cell massive MIMO system, under which the covariance approach has the key advantage that it is capable of detecting a much larger number of active devices. This thesis focuses on the question: given a signature sequence length and a set of non-orthogonal signature sequences, how many simultaneously active devices can be identified when the number of antennas at the BS is large. To answer this question, this thesis derives a necessary and sufficient condition on the Fisher information matrix associated with the estimation of effective channel statistics, based on which a numerical phase transition characterization is obtained.
Source: TSpace (University of Toronto)
Author: Zhilin Chen
Institution: University of Toronto
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3140381784
