| Summary: | 博士 === 國立臺灣大學 === 電信工程學研究所 === 102 === In this thesis, a hybrid multiple access channel (MAC) between sensor to fusion center
(FC) links in wireless sensor networks (WSNs) is proposed, and the corresponding distributed
estimation and detection problems are investigated. The main contributions of this thesis are
described in the following.
In Chapter 2, the distributed estimation of a scalar signal by a power-constrained WSN is
investigated where the amplify-and-forward sensor signals are relayed to a FC over a hybrid
MAC. The proposed hybrid MAC is a composite of conventional orthogonal and coherent
MACs, and shown to offer more performance choices than both conventional channels in
terms of mean-squared error (MSE) performance and system outage probability (SOP). Both
random and deterministic scalar sources are considered for the hybrid MAC WSN in conjunction
with linear minimum MSE (MMSE) and maximum likelihood (ML) estimations,
respectively. Specifically, the optimal and suboptimal schemes are developed to allocate sensor
power under a total power constraint and these power allocation schemes are found similar
in terms of algorithm and achievable estimation performance for both scalar sources. Particularly,
closed-form expressions for sensor power are obtained for suboptimal power allocation
and two-step procedures are proposed to solve the optimal power allocation problems numerically.
Furthermore, asymptotic expressions of MSE under various limiting parametric cases
and SOP are analytically derived for suboptimal power allocation under a uniform sensor
grouping strategy. It is analytically shown that, under a uniform sensor grouping strategy, the
suboptimal power allocation scheme can offer asymptotically the same MSE performance as
the optimal power allocation scheme when the number of sensors transmitting through the
same coherent channel increases to infinity. Simulation results also show that the suboptimal
power allocation scheme provides nearly optimal MSE performance for arbitrary sensor
grouping strategies and the uniform sensor grouping is virtually the best grouping strategy.
i
Moreover, for uniformly grouped hybrid MAC, the SOP for suboptimal power allocation is
shown to provide a diversity order equal to the number of orthogonal channels when there
are sufficiently many orthogonal channels.
The distributed estimation of a deterministic scalar source is considered in Chapter 3 by a
power-constrained WSN with a FC where the phase-modulated sensor signals are relayed to
the FC over a uniformly grouped hybridMAC. In this case, ML estimate for ideal sensing environments
and a two-step estimate are developed for the estimation of source parameter and
the corresponding modified Cram’
|
|---|
