In-Networks Spatial Query Estimation in Sensor Networks

Recently technology developments enable the production and deployment of sensor networks that provide outstanding ability to monitor discrete and continuous phenomena in physical space. From a database perspective, a sensor network can be seen as a virtual distributed database system (DBS) with sensor nodes that are tiny DBS themselves. Thus, a user can interact with the sensor network as a whole, and send queries to it. The sensor nodes run tiny footprint DBS locally, and participate in global query execution. Nowadays, spatial queries over sensor network mostly retrieve discrete information measured at the location of sensor nodes. In environmental applications, however, the estimation of continuous phenomenon such as a toxic cloud or a temperature field is of great interest. Thus, strategies need to be available to evaluate spatial queries about continuous phenomena, and estimate results based on discrete sensor measurements. Kriging is a traditional spatial interpolation method with a guaranteed minimal estimation error, and a smooth, unbiased estimation. However, its execution is computationally expensive. In this paper, we present an approach to Kriging that allows inexpensive in-network execution to evaluate queries over a sensor network. Our approach, called QUAKE, optimizes the selection of sample size and sensor nodes for the requested estimation value of any point location in the sensor network.