Data Uncertainty Management in Sensor Networks

Abstract

Data readings collected from sensors are often imprecise. The uncertainty in the data can arise from multiple sources, including measurement errors due to the sensing instrument and discrete sampling of the measurements. For some applications, ignoring the imprecision in the data is acceptable, since the range of the possible values is small enough not to significantly affect the results. However, for others it is necessary for the sensor database to record the imprecision and also to take it into account when processing the sensor data. This is a relatively new area for sensor data management. Handling the uncertainty in the data raises challenges in almost all aspects of data management. This includes modeling, semantics, query operators and types, efficient execution, and user interfaces. Probabilistic models have been proposed for handling the uncertainty. Under these models, data values that would normally be single values are transformed into groups of data values or even intervals of possible values.