Correlation-aware Resource Allocation
Oct 06, 2011
from 12:00 PM to 01:00 PM
|Where||Maxwell Room, Engr. IV 57-124|
|Add event to calendar||
Advisor: Gregory J. Pottie
When several nodes in a network share limited communication resources, the media access protocol is implemented in order to determine the allocation of resources to each node. This strategy is a key component in design of networks with high transmission efficiency. Additionally, when nodes observe and transmit correlated information, their correlation characteristics can be used in the media access protocol in order to increase the efficiency of the communication network. This thesis studies the problem of cross-layer resource allocation for correlated sources. For several specific cases of single cell and multi-cell networks, novel and practicable solutions are proposed and verified with simulations.
First for the uplink transmission in a single cell, a Code Division Multiple Access (CDMA) network of correlated video sensors is considered. A novel cross- layer resource allocation strategy is proposed with the goal of maximizing the weighted average of all reconstructed video qualities. The algorithm finds the power and orthogonal code assignment to each sensor, using no communication among sensors and minimal communication between sources and the receiver. Compared with independent
methods, the cross-layer correlation-aware resource allocation achieves significant gain in average sensor video quality. Secondly, a cross-layer resource allocation strategy is proposed for multi-cell Orthogonal Frequency Division Multiple Access (OFDMA) networks of general correlated sources. This method assumes a distance based correlation model among the sources. The goal is to find the power and orthogonal frequency sub- band assignment in order to minimize the maximum distortion
achieved by any source in the network. The challenge in this case is to take both the inter-cell interference and the source correlation characteristics into consideration in the resource allocation strategy. Our proposed solution solves this large NP-hard problem in three simple, workable steps.
Additionally, this thesis contributes to the problem of resource allocation for general multi-cell OFDMA communication networks by introducing a novel inter-cell interference management scheme, called ICon. This method is based on concentrating the interference to a cell on a pre-determined frequency band, which is adapted in order to balance the performance across the network.
Dorna Bandari is a Ph.D. Candidate in Electrical Engineering. She is currently working in the Center for Embedded Networked Sensing (CENS) at the University of California, Los Angeles (UCLA). Dorna attended the UCLA for her undergraduate education where she received a B.S. degree with Summa Cum Laude in Electrical Engineering with an emphasis on Signal Processing. Her M.S. degree is also from UCLA in Electrical Engineering in Signal Processing with specialization in optimization.
Dorna has held an internship at the Wireless Connectivity group at Broadcom Corp., and was a guest PhD student at the Signal Processing Laboratory (LTS4) at École Polytechnique Fédérale de Lausanne (EPFL) from 2009 to 2011.
Her research interests lie in wireless communication multiple access protocols and adaptive resource allocation.