Session S6

Session S6: Wireless Sensor Networks

Time:                 Tuesday, May 13, 14:30-15:45
Chair:                 Cihan Tepedelenlioglu, Arizona State University
Co-Chair:           Sergio Cabrera, University of Texas at El Paso

S6-1: Optimal Recovery for Non-Uniform Signal Interpolation: a Re-Visit for Application to Wireless Sensor Networks
Sergio Cabrera, Lata Upadhyayula and Jose Gerardo Rosiles, University of Texas at El Paso, El Paso TX
In this paper, we investigate the basic problem of recovery of signals from non-uniform samples based on the framework of optimal recovery in Hilbert spaces. The method is illustrated with 1-D and 2-D signals for irregular sampling configurations that are relevant in Wireless Sensor Networks applications. The optimal approach discussed here can serve as a baseline for comparison against techniques that are being developed taking into account the constraints imposed by the WSN application, In contrast to spline interpolation, the optimal recovery method allows the user to incorporate bandwidth or spectral shape information in the recovery process. Examples are presented using 1-D test signals and comparisons are made with the cubic spline approach. The extension to 2-D signals is discussed and illustrated for a spatio-temporal signal non-separable spectral support.

S6-2: Distributed Estimation: Do Not Trust Gossips
Carlos Mosquera University of Vigo, Galicia, Spain and Sudharman Jayaweera University of New Mexico, Albuquerque NM
Distributed estimation and detection is of interest for those situations for which the sensor net must achieve an agreement by exchanging information without resorting to the use of an external fusion center. In this paper we deal with the distributed estimation of a parameter for both static and time-varying cases, for which it is important to have similar estimates as accurate as possible. The cooperation is performed in a distributed way to guarantee scalability and robustness to failures, and it is designed to reduce the detrimental effects of the channel noise on the sensor exchanges.

S6-3: Cooperative Diversity with Continuous Phase Modulation
Anna-Marie Silvester, Lutz Lampe and Robert Schober University of British Columbia, Canada
Cooperation among wireless devices has been shown to improve the error performance and capacity of wireless systems. However, when operating in wireless sensor networks, cooperating devices face a number of unique challenges including a stringent limitation on power consumption due to limited battery capacity. To date, the distributed space-time (ST) codes that have been proposed to enable node cooperation use only linear modulation that incur high peak-to-average power ratios (PAPRs) and result in increased power consumption in linear power amplifiers. In such energy-constrained conditions, continuous-phase modulation (CPM) could be considered a natural choice because of its constant-envelope property that enables the use of not only energy-efficient, but also inexpensive nonlinear power amplifiers. Therefore, we propose distributed ST codes that employ CPM. The proposed codes are designed to operate in wireless networks containing a large set of nodes N, of which only a small a priori unknown subset S_N will be active at any time. We adopt diagonal block-based ST codes and consider the optimization of two classes of signature vectors: a general set that fixes the envelope of blocks of data but allows the signal envelope to vary from block to block, and a “constant-envelope set” that constrains a node to maintain constant envelope over all blocks it transmits while relaying a given signal. We provide appropriate design rules for signature vector construction, and efficient numerical methods for the generation of signature sets. Through the combination of ST-CPM codes and signature vector sets, the proposed distributed ST-CPM codes allow for energy-efficient cooperative transmission and low complexity implementations.

S6-4: Distributed Estimation for Cognitive Radio Networks - the Binary Symmetric Channel Case
Murat Senel, Vibhav Kapnadak and Edward J. Coyle Purdue University, West Lafayette IN
A cognitive radio network characterizes incumbents’ radios in a slice of RF spectrum to determine if it can use it without causing/suffering interference. We consider a case of distributed estimation in which wireless sensors measure the SNR of an incumbent and communicate their measurements to a clusterhead. Each sensor sends only one bit, ±1, indicating whether the SNR is greater or less than that sensor’s threshold. The cluster head (CH) fuses these measurements to produce a maximum-likelihood estimate (MLE) of the distance to the incumbent. Both the measurement and communication tasks are affected by noise. In this paper, we assume a Binary Symmetric Channel (BSC) between the sensors and the CH and derive the MLE when the sensors’ thresholds are different – unlike prior work in which all sensors used the same threshold. We investigate the performance of our scheme via simulations, derive the Cramer-Rao lower bound for the MLE, and compare our scheme with prior work. These investigations show that our MLE performs as well as the MLE when every node uses the same optimal threshold. Our scheme performs better when the single threshold deviates from the typically unknown optimal value.

S6-5: Multi-Stage Data Fusion for Distributed Detection
Yingwei Yao University of Illinois at Chicago, Chicago IL
We propose a multi-stage data fusion scheme for distributed detection in wireless sensor networks. This scheme groups sensors according to the informativeness of their data. Fusion center collects sensor data sequentially, starting from the most informative data and terminates the process when the target performance is reached. We analyze the average sample number needed in this scheme. Our analysis, verified by simulations, shows that the proposed scheme significantly reduces the communication cost.

S6-6: Consensus-based k-means Algorithm for Distributed Learning using Wireless Sensor Networks
Pedro A. Forero, Alfonso Cano and Georgios Giannakis University of Minnesota, Minneapolis MN
The k-means algorithm is a popular approach to various machine learning applications where data are typically available at a central location. The present paper develops a decentralized k-means algorithm to estimate the location of k centers that identify clusters in data collected at spatially deployed wireless sensors. The membership assignment step in the novel iterative scheme relies on local information available to individual sensors, while during the center update step sensors exchange information only with their one-hop neighbors to reach consensus and eventually percolate the global information needed. Analysis and simulations demonstrate that the resultant consensus-based distributed k means (CBDk-means) algorithm matches well the resource-limited characteristics of wireless sensor networks and compares favorably with existing alternatives because it has wider applicability and remains resilient to inter-sensor communication noise.

S6-7: Space-Time Reversal Techniques for Information Retrieval in Wireless Sensor Networks
Thiagarajan Sivanadyan and Akbar Sayeed University of Wisconsin-Madison, Madison WI
In this paper, we explore the benefits of space-time reversal (STR) techniques in the context of information retrieval in wireless sensor networks under the recently proposed Active Wireless Sensing (AWS) framework. In AWS, individual sensors are differentiated via their distinct space-time signatures and STR schemes exploit these sensor signatures. In the downlink, STR techniques are proposed to address individual sensors thus enabling sensor programming for different information retrieval or sensing modes. In the uplink, we propose an STR scheme for localizing distinct sensor responses to distinct signal space dimensions at the WIR thereby reducing interference between sensor transmissions. Furthermore, it does not require explicit estimation of sensor signatures at the WIR. The benefits of STR techniques in both scenarios are quantified analytically and illustrated with physically meaningful simulation results.