In this thesis the problem of distributed detection, localization, and estimation
(DDLE) of a stationary target in a fusion center (FC) based wireless sensor network
(WSN) is considered. The communication process is subject to time-critical
operation, restricted power and bandwidth (BW) resources operating over a shared
communication channel Buffering from Rayleigh fading and phase noise. A novel algorithm
is proposed to solve the DDLE problem consisting of two dependent stages:
distributed detection and distributed estimation. The WSN performs distributed
detection first and based on the global detection decision the distributed estimation
stage is performed. The communication between the SNs and the FC occurs over a
shared channel via a slotted Aloha MAC protocol to conserve BW.
In distributed detection, hard decision fusion is adopted, using the counting
rule (CR), and sensor censoring in order to save power and BW. The effect of
Rayleigh fading on distributed detection is also considered and accounted for by
using distributed diversity combining techniques where the diversity combining is
among the sensor nodes (SNs) in lieu of having the processing done at the FC.
Two distributed techniques are proposed: the distributed maximum ratio combining
(dMRC) and the distributed Equal Gain Combining (dEGC). Both techniques show
superior detection performance when compared to conventional diversity combining
procedures that take place at the FC.
In distributed estimation, the segmented distributed localization and estimation
(SDLE) framework is proposed. The SDLE enables efficient power and BW
processing. The SOLE hinges on the idea of introducing intermediate parameters
that are estimated locally by the SNs and transmitted to the FC instead of the
actual measurements. This concept decouples the main problem into a simpler set
of local estimation problems solved at the SNs and a global estimation problem
solved at the FC. Two algorithms are proposed for solving the local problem: a
nonlinear least squares (NLS) algorithm using the variable projection (VP) method
and a simpler gird search (GS) method. Also, Four algorithms are proposed to solve
the global problem: NLS, GS, hyperspherical intersection method (HSI), and robust
hyperspherical intersection (RHSI) method. Thus, the SDLE can be solved through
local and global algorithm combinations. Five combinations are tied: NLS2 (NLS-NLS),
NLS-HSI, NLS-RHSI, GS2, and GS-N LS. It turns out that the last algorithm
combination delivers the best localization and estimation performance. In fact , the
target can be localized with less than one meter error.
The SNs send their local estimates to the FC over a shared channel using the
slotted-Aloha MAC protocol, which suits WSNs since it requires only one channel.
However, Aloha is known for its relatively high medium access or contention delay
given the medium access probability is poorly chosen. This fact significantly
hinders the time-critical operation of the system. Hence, multi-packet reception
(MPR) is used with slotted Aloha protocol, in which several channels are used for
contention. The contention delay is analyzed for slotted Aloha with and without
MPR. More specifically, the mean and variance have been analytically computed
and the contention delay distribution is approximated. Having theoretical expressions
for the contention delay statistics enables optimizing both the medium access
probability and the number of MPR channels in order to strike a trade-off between
delay performance and complexity.