Publication

Abstract : We examine the problem of tracking dynamic boundaries occurring in natural phenomena using sensor networks. Remotelyplaced sensor nodes produce noisy measurements of various points on the boundary using range-sensing. Two main challengesof the boundary tracking problem are energy-efficient boundary estimations from noisy observations and continuous tracking ofthe boundary. We propose a novel approach which uses discrete estimations of points on the boundary using a regression-basedspatial estimation technique and a smoothing interpolation scheme to estimate a confidence band around the entire boundary. Inaddition, a Kalman Filter-based temporal estimation is used to help selectively refresh the estimated boundary at a point onlyif the boundary is predicted to move out of the previous estimated intervals at that point. An algorithm for dynamic boundarytracking (DBTR), the combination of temporal estimation with an aperiodically updated spatial estimation, allows us to providea low overhead solution to track dynamic boundaries that does not require prior knowledge about the nature of the dynamics.Experimental results demonstrate the effectiveness of our algorithm and estimated confidence bands achieve loss of coverage ofless than 2% for smooth boundaries.