Energy-Efficient Map Interpolation for Sensor Fields Using Kriging

We propose a spatial autocorrelation aware, energy efficient, and error bounded framework for interpolating maps from sensor fields. Specifically, we propose an iterative reporting framework that utilizes spatial interpolation models to reduce communication costs and enforce error control. The framework employs a simple and low overhead in-network coordination among sensors for selecting reporting sensors so that the coordination overhead does not eclipse the communication savings. Due to the probabilistic nature of the first round reporting, the framework is less sensitive to sensor failures and guarantees an error bound for all functional sensors for each epoch. We then propose a graceful integration of temporal data suppression models with our framework. This allows an adaptive utilization of spatial or temporal autocorrelation based on whichever is stronger in different regions of the sensor field. We conducted extensive experiments using data from a real-world sensor network deployment and a large Asian temperature dataset to show that the proposed framework significantly reduces messaging costs and is more resilient to sensor failures. We also implemented our proposed algorithms on a sensor network of MICAz motes. The results show that our algorithms save significant energy and the out of bound errors due to packet loss are below 5%.