The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Probabilistic approaches to the design of wireless ad hoc and sensor networks
Author:	Guo, Zhen
View Online:	njit-etd2006-026 (xiii, 83 pages ~ 3.6 MB pdf)
Department:	Department of Electrical and Computer Engineering
Degree:	Doctor of Philosophy
Program:	Computer Engineering
Document Type:	Dissertation
Advisory Committee:	Zhou, MengChu (Committee chair) Ge, Hongya (Committee member) Hu, Jie (Committee member) De, Swades K. (Committee member) Shih, Frank Y. (Committee member)
Date:	2006-01
Keywords:	Wireless networks
Availability:	Unrestricted
Abstract:	The emerging wireless technologies has made ubiquitous wireless access a reality and enabled wireless systems to support a large variety of applications. Since the wireless self-configuring networks do not require infrastructure and promise greater flexibility and better coverage, wireless ad hoc and sensor networks have been under intensive research. It is believed that wireless ad hoc and sensor networks can become as important as the Internet. Just as the Internet allows access to digital information anywhere, ad hoc and sensor networks will provide remote interaction with the physical world. Dynamics of the object distribution is one of the most important features of the wireless ad hoc and sensor networks. This dissertation deals with several interesting estimation and optimization problems on the dynamical features of ad hoc and sensor networks. Many demands in application, such as reliability, power efficiency and sensor deployment, of wireless ad hoc and sensor network can be improved by mobility estimation and/or prediction. In this dissertation, we study several random mobility models, present a mobility prediction methodology, which relies on the analysis of the moving patterns of the mobile objects. Through estimating the future movement of objects and analyzing the tradeoff between the estimation cost and the quality of reliability, the optimization of tracking interval for sensor networks is presented. Based on the observation on the location and movement of objects, an optimal sensor placement algorithm is proposed by adaptively learn the dynamical object distribution. Moreover, dynamical boundary of mass objects monitored in a sensor network can be estimated based on the unsupervised learning of the distribution density of objects. In order to provide an accurate estimation of mobile objects, we first study several popular mobility models. Based on these models, we present some mobility prediction algorithms accordingly, which are capable of predicting the moving trajectory of objects in the future. In wireless self-configuring networks, an accurate estimation algorithm allows for improving the link reliability, power efficiency, reducing the traffic delay and optimizing the sensor deployment. The effects of estimation accuracy on the reliability and the power consumption have been studied and analyzed. A new methodology is proposed to optimize the reliability and power efficiency by balancing the trade-off between the quality of performance and estimation cost. By estimating and predicting the mass objects' location and movement, the proposed sensor placement algorithm demonstrates a siguificant improvement on the detection of mass objects with nearmaximal detection accuracy. Quantitative analysis on the effects of mobility estimation and prediction on the accuracy of detection by sensor networks can be conducted with recursive EM algorithms. The future work includes the deployment of the proposed concepts and algorithms into real-world ad hoc and sensor networks.