Using an evolution model for efficient estimation and tracking of dynamic boundaries

We address the reduction of communication costs in contour tracking sensor networks. Sensors sample points on the contour. The sample points are modeled as evolving according to Brownian motion with constant drift. The model is then used by the sink to estimate the sample points on the contour in lieu of actual measurements. The Brownian motion model allows the sink to also obtain the confidence in its estimate. The sensors are queried for new measurements when the confidence in the estimates is below a specified threshold. In addition to new measurement, the sensors also report drift and variance parameter samples. The theory for the model and the update algorithm is developed, the estimation algorithm is described, and the method is analyzed using simple simulation models. Results indicate a promising way to efficiently monitor contours in dynamic environments.