"Sensor Node Localization Based on Two-Way Time-of-Arrival Ranging - TopicsExpress

"Sensor Node Localization Based on Two-Way Time-of-Arrival Ranging with Imperfect Clocks" Wed, Aug 28, 2013 @ 02:00 PM - 03:00 PM EEB 248 Erik Strom, Chalmers University of Technology Abstract: In this talk, we will discuss the positioning of sensors nodes based on two-way time-of-arrival (TW-TOA) measurements when the nodes have imperfect clocks. More precisely, the problem is to localize a single target node using distance measurements to a number of nodes at a priori known positions (anchor nodes). The target node clock is assumed to follow an affine relationship with the anchor node clocks. That is, the target node clock will, in general, run too quick or too slow and will be offset compared with the anchor node clocks. The clock rate is sometimes called clock skew and is ideally equal to one, indicating that the clock runs at the same rate as the global reference time. The anchor nodes are assumed to have perfect clock skews, but unknown and different offsets. The TW-TOA measurement process will remove the clock offsets, but the clock skew difference will affect the distance measurements and, therefore, also the position estimate, if not properly accounted for. We model the target node clock skew as a nuisance parameter and show that the resulting maximum likelihood (ML) estimator is difficult to compute. To find more tractable estimators, we apply a nonlinear pre-processing step to convert the ML problem into a linear least squares problem under a quadratic constraint. The latter problem is shown to be a special case of the so-called generalized trust region problem, which we can solve exactly under mild conditions. We develop two suboptimal positioning methods and compare the performance and complexity with the ML estimator and the Cramer-Rao bound. The developed methods are numerically shown to offer good performance, but with less complexity (and accuracy) compared with the ML estimator. This presented research is joint work with Mohammad Gholami, Chalmers University, and Sinan Gezici, Bilkent University.