Center for Embedded Network Sensing

Today numerous sensor networks are deployed in various environments and collecting sensor data. However, most sensor networks are isolated. Each uses customized methods to deliver data to their users and the data sharing are not considered. We introduce a sensor data republishing that consists of source data retrieving, data processing and publishing back the transformed data. It is difficult for other users to process further on existing published sensor data such as aggregation, filtering, statistical estimation, vetting and error suppression. Especially when source data are coming from several different sources, it becomes more complicated to manage and republish the transformed data. We are exploring what kind of software infrastructure is required for sensor data republishing. Currently we are looking at how to provide data providence (tracking source data) with minimal storage, how to tolerate occasional network outages and how republishers can learn and adapt to vary data generation rates. As a preliminary result, we propose a data link that interconnects between transformed data and source data. We also propose an incremental compression algorithm to store sensor data links efficiently. Our republishing mechanisms will extend SensorBase.org.