Sentient computing : a universal framework for spatial data

• Main Author: Fawcett, J. K.
• Published: University of Cambridge 2004
• Subjects: 006.3
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.598965

Spatial datasets are an important source of context for computer applications in Sentient Computing. However, heterogeneous sensor networks present location and orientation telemetry in differing formats, with incompatible interpretation semantics, and subject to contrasting measurement and quantisation errors. Applications consume data in another set of frames of reference and are generally confined to operating in rooms and office spaces that are sensed by a single, homogeneous suite of sensors under the same ownership and administrative control. This thesis presents Scalable, Distributed Spatial Indexing (SDSI), a unifying framework designed to capture the semantics of sensor technologies and autonomously transform, re-write, and otherwise massage data so as to interface to application-level software and decouple applications and sensor hardwares to support seamless mobility. Translation Functions perform the mathematical operations necessary to convert raw sensor data to other formats. The SDSI framework ensures the preservation of the mathematical and scientific rigour associated with the interpretation of each telemetry data stream. Uncertainties and errors are also converted by Translation Functions and allow client applications to determine the accuracy and precision suitability of any sensor data for individual purposes. SDSI builds on long-standing database theory to provide security and privacy controls, allowing users to restrict access to sensitive data streams without requiring multiple, isolated data handling systems. Caching techniques are used to maintain processing throughput and network distribution provides scalable, inter-site communication through the global Internet. The evaluation includes a substantial investigation of Driveable Computing—the scope for Sentient Computing in private motor vehicles—in which spatial frames of reference are required to interpret cartographic data, interact with human occupants using natural language, interface with positioning hardware, and so forth. The conclusion, that the proposed framework is suitable and applicable to Sentient Computing, is extremely positive and poses several avenues and opportunities for continued research in the field.