Semantic Challenges in Sensor Networks

Motivation

We have seen significant recent progress in quantity and capabilities of mobile devices with the ability to perform human-in-the-loop sensing, wireless sensors and sensor networks. These, combined with improved ability to bridge the physical and cyber world divide, have fostered the broad availability of sensor data capturing the state of the physical world. Promising and already successful examples are applications in environmental monitoring, agriculture, surveillance and intrusion detection, public security, and supply chain management. Furthermore, ideas towards a Web of sensors have been proposed which is to be understood as a (large scale) network of spatially distributed sensors. Also, terms like "Internet of Things", "Collaborating Objects" and "Ambient Intelligence" emphasize the trend to a tighter connection between information systems and the physical world.

The existence of a huge number of sensor sources producing data continuously results in tremendous data volumes which are often valid or useful only for certain period of time and are never inspected by humans. In order to make sensor data useful despite the lack of human supervision in the loop, semantic annotation and analysis becomes a key component in setting up sensor data-based applications: Only if sensors and sensor data are annotated and enriched by information describing their meanings, source, and validity scope, they can be automatically discovered, processed and combined with other data in an open world. The kind of useful semantics ranges from technical metadata describing the sensors and the measurements (time, location, sensor type, validity, measurement error etc. as partially captured by standardization proposals like SensorML) to emergent semantics derived by aggregating, combining, analyzing, and enriching the raw data, e.g., in the form of analytical models, annotations, correlations etc. on the other spectrum, the data collected by human-in-the-loop sensing is small but of significant verity and complexity (e.g., language nuances, and capturing sentiments and emotions), which offer additional challenges to annotation, integration and analysis of such data.

Modeling, representing, discovering and deriving as well as using semantics for sensor data raise several challenges which are related to different aspects of developing, deploying, and using sensor network based applications. Thus, the goal of this seminar is to bring together researchers from relevant areas like:

• sensor node providers and sensor networking,
• data fusion and data stream processing,
• sensor middleware,
• geospatial and uncertain data management,
• semantic integration and Semantic Web, and
• social computing and collective intelligence.

In all these areas semantics plays an important role, either during producing and enriching data with explicit semantics or by exploiting semantics for data processing. Sharing and exchanging knowledge and experiences could result in significant synergy effects.

The seminar will focus on the following major issues:

• methodologies and languages for modeling and representation, issues of sensing-perception-semantics,
• standards, ontologies, and middleware for semantic sensor networks, semantic annotation of high throughput machine sensor data as well as social/human-in-the-loop sensing data,
• emergent semantics in sensor networks and sensor data processing,
• exploiting sensor data semantics for geospatial and uncertainty data management, and
• specific use cases and applications of semantic sensor networks.

The intention is to analyze the state of the art in the different areas with respect to semantics, discuss problems, specific methodologies and applications of semantic-aware sensor networks and emergent semantics as well as to identify future trends and research directions.