Date of Graduation
Statler College of Engineering and Mineral Resources
Lane Department of Computer Science and Electrical Engineering
Brian D. Woerner.
The area of wireless sensor networks (WSN) is a fast growing area and has a variety of applications such as environmental monitoring, health care, agricultural applications, and military applications. While this research area has been active since 1980s, new applications bring new constraints, and these lead to new open problems. This thesis approaches one of them. It focuses on the problem of target location estimation with limited knowledge about sensor locations in a WSN. Sensors in a WSN are randomly deployed over a known area. The true locations of the sensors are unknown, but each sensor node is equipped with positioning technolgy system such as GPS generating noisy measurements of the true unknown location. A target is modeled as a point source generating a spatial parametric signal or field. The shape of the parametric field is known, but the location of the point source is unknown. The sensors monitor the field and report their noisy measurements to a base station called Fusion Center (FC). The measurements are wirelessly transmitted to the FC via Additive White Gaussian Noise (AWGN) channels. We assume two cases of transmission channel in this thesis. In the first case, the observed data are modulated using a linear analog modulation such as amplitude. In the second case, the observed data are quantized to M quantization levels and then transmitted to the FC using a digital modulation scheme such as ON-OFF Keying (OOK). Each sensor transmits two sets of data: (1) noisy measurements of the field and (2) the measurements of the sensor position provided by its positioning system. Given noisy measurements of the field and noisy measurements of sensor locations, the task of the FC is to find the location of the point source. The FC applies the Maximum Likelihood estimation approach to solve for unknown parameters. In this thesis, the numerical solution is due to the Bisection method that iteratively estimates the location of the point source and the location of sensors and alternates these two steps. The field generated by a point source is modeled as a Gaussian bell function. The measurements of sensor location are assumed to be drawn from a two-dimensional Gaussian distribution with the true location of a sensor as its mean and known covariance matrix. The performance of the proposed solution is measured in terms of the square error (SE) between the true and estimated location of the point source. The MS is analyzed as a function of many parameters of the WSN, FC and parametric field. A comparison with a baseline case, when the locations of sensors are known to the FC, is made.
Ali, Muammar, "Distributed Joint Sensor and Target Location Estimation" (2013). Graduate Theses, Dissertations, and Problem Reports. 3605.