Distributed estimation and detection in wireless sensor networks
Abstract
Recent advances in wireless sensors, communications and MEMS technology have enabled large scale deployments of low cost Wireless Sensor Networks (WSN) for various practical applications; for example in smart power grid networks, for vehicular communication networks, design of energy efficient buildings and in health-care areas. While designing WSNs for any application we face two important challenges. The first challenge is to develop energy-efficient distributed signal processing algorithms that can perform accurate sensing of the underlying physical phenomenon in the monitoring environment, and the second is to determine the optimal deployment of wireless sensors for a particular sensing application. In this thesis, we address these issues by proposing energy-efficient distributed signal processing algorithms and determining optimal deployment schemes to solve estimation and detection problems in WSNs. We first consider the problem of estimating the location of an interferer operating in the 2.4 GHz ISM band using a cluster of wireless sensor motes. We propose a low-complexity 1-bit quantization framework and compare the performance over the BSC and the orthogonal multiple-access fading channel of the Maximum- Likelihood Estimator (MLE) with other techniques proposed in the literature. We then propose an iterative quantization technique, wherein limited feedback from the Cluster Head (CH) is used to adjust the quantization thresholds at the sensor motes. We prove asymptotic properties of the MLE obtained using our scheme and via simulation show that our proposed technique is more accurate and adapts faster than previous approaches. Next we study the asymptotic performance of distributed detection algorithms in clustered multi-hop wireless sensor networks for both fixed and randomly deployed networks. We derive the error exponents to characterize the asymptotic performance of the Decision Error Probability (DEP) at the CH and study the effects on the DEP due to adding extra rings and increasing the dimensionality of the network. Finally we study the optimal deployment schemes for WSNs. Unlike previous approaches that aim optimize energy costs or lifetime, we consider a different approach of deriving the optimal node density requirements that minimize the DEP at the CH subject to network lifetime constraints. We formulate the optimization problem taking into all factors such as the underlying MAC protocol used, defective sensor motes and signal propagation characteristics. We derive the optimal node densities for both single and multi-ring single hop cluster networks and show the trade-offs between energy used, DEP and the number of motes deployed.
Degree
Ph.D.
Advisors
Bell, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.