Signal Processing In Radar And Non-radar Sensor Networks
MetadataShow full item record
This dissertation studies six topics within the area of radar andnon-radar sensor networks from a signal processing perspective:radar sensor networks (RSN) waveform design and performance analysis(chapter 2), blind speed alleviation using RSN (chapter 3), targetdetection in foliage using Ultra-Wideband (UWB) RSN (chapter 4),sense-through-foliage\&wall channel modeling (chapter 5), channelselection algorithms in virtual multiple-input-multiple-output(MIMO) sensor networks (chapter 6) and RF emitter passivegeolocation using unmanned aerial vehicles (UAVs) and sensors(chapter 7).In RSN, distributed radar sensors work in an ad hoc fashion but aregrouped together by an intelligent clusterhead that combineswaveform diversity. RSN not only provide spatial resilience fortarget detection and tracking compared to traditional radars, butalso alleviate inherent radar defects such as the blind speedproblem. This interdisciplinary area offers a new paradigm forsignal processing research. In this dissertation, orthogonalconstant frequency (CF) pulse waveforms are designed for bothcoherent and noncoherent RSN detection systems. To what extend RSNoutperform single radar and how Doppler shift degrades theperformance are analyzed in terms of probability of detection andprobability of false alarm.As blind speed problem can turn out to be a catastrophe to movingtarget detection, RSN design with equal gain combination (EGC)algorithm is proposed to tremendously alleviate this problem. Afuzzy logic system (FLS) is also designed to optimize the number ofradars in RSN, making the FLS-based RSN achieve somehow constantprobability of miss detection even with different systemconfiguration.In foliage, UWB RSN are employed for target detection. On a basis ofpragmatic measurements, a RSN Rake structure and two signalprocessing schemes are proposed to improve the target detectionperformance. One is differential-based approach that accounts forthe channel effect and analyzes the ``defoliated'' signal. Anotherapplies short-time Fourier transform (STFT) that uses a slide windowto determine the sinusoidal frequency and phase content. Bothschemes are able to detect the target successfully.Based on these real radar data, new sense-through-foliage channelmodel is proposed and parameters are statistically analyzed. Theamplitude can be characterized by log-logistic distribution whilethe time arrival of multi-path contributions can be modeled as aPoisson process. Another statistical model for sense-through-wallchannel is also proposed based on experimental measurement using UWBnoise radar. These results provide an improved understanding ofwireless propagation in foliage and wall.In non-radar virtual MIMO wireless sensor networks (WSN), twopractical algorithms to select a subset of channels are presented tobalance the MIMO advantage and the energy consumption of sensorcooperation. If intra-cluster node-to-node multi-hop needs be takeninto account, Maximum Spanning Tree Searching (MASTS) algorithm inrespect of cross-layer design always provides a path connecting allsensors. When WSN is organized in a manner of cluster-to-clustermulti-hop, Singular-Value Decomposition-QR with Threshold (SVD-QR-T)approach selects the best subset of transmitters while keeping allreceivers active. Simulations show that both algorithms providesatisfying performances with reduced resource consumption.Finally, a network of UAVs is designed for passive location of RFemitters. Each UAV is equipped with multiple electronic surveillance(ES) sensors to provide local mean distance estimation based onreceived signal strength indicator (RSSI). Fusion center willdetermine the location of the target through UAV triangulation.Different with previous existing studies, this method is on a basisof an empirical path loss and log-normal shadowing model, from awireless communication and signal processing vision to offer aneffective solution.