This chapter considers the sensor network localization problem using signal strength. Signal strength information is stored in a kernel matrix. Least squares kernel regression methods are then used to get an estimate of the location of unknown sensors. Locations are represented as complex numbers with the estimate function consisting of a linear weighted sum of kernel entries. The regression estimates have similar performance as previous localization methods using kernel classification methods, but at reduced complexity. Simulations are conducted to test the performance of the least squares kernel regression algorithm. We also consider the cases where sensors are mobile and on-line kernel regression learning algorithms are formulated to track moving sensors. Finally, we discuss some physical constraints on the sensor networks (i.e., communication and power constraints). To deal with these constraints, we proposed using distributed learning algorithms to cut down on communications between sensors. An ensemble of learners each solve a kernel regression algorithm and then communicate among each other to reach a solution. The communication costs are lowered using distributed learning algorithms and through simulations we show that the performance is comparable to the centralized kernel regression solution. © 2008 Springer US.
CITATION STYLE
Kuh, A., & Zhu, C. (2008). Sensor network localization using least squares kernel regression. In Signal Processing Techniques for Knowledge Extraction and Information Fusion (pp. 77–96). Springer US. https://doi.org/10.1007/978-0-387-74367-7_5
Mendeley helps you to discover research relevant for your work.