Distributed Signal Processing and Data Fusion Methods for Large Scale Wireless Sensor Network Applications

Abstract. In this PhD dissertation we study the problem of continu-ous object tracking using large scale WSNs. We propose a novel practical WSN-based scheme that is able to track and predict the evolution behav-ior of a continuous object’s boundary under realistic assumptions. The proposed scheme consists of two main components: a) A collaborative in-network WSN algorithm that estimates the local evolution parameters (orientation, direction and speed) of an evolving continuous object, and b) a novel algorithm which combines the produced local estimates, as they become available to a fusion center, to reconstruct the overall con-tinuous object’s boundary. Extensive computer simulations demonstrate the ability of the proposed collaborative algorithm to estimate accu-rately the evolution characteristics of complex continuous objects (e.g. with time-varying evolution rates and/or irregular boundary shapes) us-ing reasonably dense WSNs. Moreover, it shown that the algorithm is robust to sensor node failures and communication link failures which are expected in harsh environments. Finally, we show that the proposed boundary reconstruction algorithm is able to track with accuracy the evolution of different types of continuous objects, using a small number of local front estimates that may be distorted with error