Semantic Streams



Research Publications Misc
Overview One of the largest remaining obstacles to the widespread use of sensor networks is the interpretation of the sensor data; users must be able to query for high-level facts about the world, not just raw ADC readings. We designed a framework called Semantic Streams which allows users to pose declarative queries for semantic values, such as "the speeds of vehicles near the entrance of the parking garage." The system combines logical inference with AI planning techniques to compose a sequence of inference units, which are stream operators that perform a minimal amount of sensor fusion or interpretation on incoming event streams and incorporate newly inferred semantic values into outgoing event streams. Both the sensor network and the inference units are logically described in Prolog and the system can reason about which sensors in space to use and whether the inference units are being used in an appropriate world context. Typically, multiple combinations of sensors and inference units can answer the same query, so users can also declare QoS constraints in CLP(R) notation to choose between logically equivalent inference graphs, for example: "the confidence of the speed estimates should be greater than 90%." or ``I want to minimize the total number of radio messages.'' This work is patent pending.
Publications

Kamin Whitehouse, Jie Liu, Feng Zhao."Semantic Streams: a Framework for Composable Inference over Sensor Data". The Third European Workshop on Wireless Sensor Networks (EWSN), Springer-Verlag Lecture Notes in Computer Science. Zurich, Switzerland. February 13-15, 2006. (ppt)

"Automatic Programming with Semantic Streams", Kamin Whitehouse, Feng Zhao, Jie Liu. The 3rd ACM Conference on Embedded Networked Sensor Systems (SenSys '05), November 2-4 2005.



Kamin Whitehouse
Computer Science Department
The University of Virginia
217 Olsson Hall
Charlottesville, Virginia 94720