Recent advances in hardware technology have led to the emergence of small, low-power,
and possibly mobile sensors with limited onboard processing and wireless communication capabilities.
When deployed in large numbers over space, these individually primitive sensors
can cooperate to form an intelligent network, the wireless sensor network (WSN), capable
of measuring aspects or identities of the operational environment with unprecedented accuracy.
This is a promising technology ideal for applications as diverse as environmental and
healthcare monitoring, smart-house climate control, tactical surveillance, space exploration,
and intelligent transportation, to name a few. The advent of WSNs enables re-thinking the
field of distributed processing, whereby distributed sensors collaborate to perform power-
efficient tracking of nonstationary processes, and reduced-complexity detection of multiple
hypotheses. In this thesis, WSN-based distributed tracking and detection algorithms are
developed, and analyzed in terms of their optimality, robustness, as well as performance.
The underlying mobility and spatial diversity offered by WSNs gives rise to the interest
in distributed tracking of nonstationary signals, and motivates well the distributed
counterpart of Kalman filtering developed in this thesis, that is based on judicious shar-
ing of sensor observations. Different from the traditional (centralized) Kalman filter, the
low-energy budget per sensor necessitates transmission of reduced-dimensionality data and
awareness to imperfect sensor links as integral parts of the distributed design. Adhering to
these operational conditions, optimal transmission schemes are developed to minimize the
corresponding tracking error by judicious allocation of each sensor's limited power in order
to facilitate the fusion of most informative observations.
Through wireless broadcast communications, WSNs offer a suitable platform to realize
cooperative information exchange. To comply with their low-complexity radio frequency circuit,
individual sensors collaborate to eliminate the ambiguity and detect the broadcasting
message, either coded or modulated. As the number of candidate messages grows exponen-
tially, traditional distributed detection algorithms cannot operate with the sensors' limited
computation and communication capabilities. This motivates the reduced-complexity dis-
tributed decoding and demodulation algorithms of this thesis that rely on in-network one-
hop communications to achieve consensus on the sufficient statistics required to decipher the
broadcasted message. For both algorithms, the robustness to imperfect inter-sensor links
affected by additive noise or random link failures is established, and error rate analysis is
provided to evaluate their performance.