Team: Tessa Daniel, Dr Elena Gaura, Dr James Shuttleworth.
The increase in availability and affordability of wireless technology has led to a proliferation of large wireless sensor networks (WSNs) with increasing numbers of nodes deployed to resolve complex “informational” problems. Typically, however, nodes in these networks have limited resources including energy (given the limited battery power), memory, computing power and communication bandwidth.
Regardless of the application domain and deployment scope, the ability to retrieve information is critical to the successful functioning of any WSN system. In general, information extraction procedures can be categorised into three main approaches: agent-based, query-based and macro-programming. Of the three, query-based systems are the most popular mainly because they provide a usable, high level interface to the sensor network while abstracting away some of the low level details like the network topology and radio communication. In contrast, macro-programming provides a more general-purpose approach to distributed computation compared with traditional query-based approaches and focuses on programming the network as a whole rather than programming the individual devices that form the network. The agent-based approach tailors the information extraction mechanism to the type of information needed and the configuration of the network it needs to be extracted from.
In this project, a hybrid approach to information extraction is developed that aims to retain the simplicity and ease of use of query-based approaches while allowing the inclusion of useful logical abstractions provided by macro-programming approaches, to facilitate construction and resolution of more powerful end-user queries. This approach also incorporates some of the principles of agent-based systems, such as collaboration and in-network decision making, to assist in query resolution. Following an extensive survey of the literature in the area of WSN information extraction, the work to-date has highlighted the benefits of innetwork processing. The feasibility of the approach has been demonstrated through simulation.
|
|
|
|
As well as information extraction, a second step in gaining understanding of observed phenomena and exploiting the foremost potential of WSNs (that of identifying spatio-temporal patterns) is through visualisation of extracted information. Cogent’s approach is to produce multi-dimensional field representations of sensed data through interpolation techniques and algorithms. Additional aims are to provide the user with tool sets for extraction of features under observation at a level of precision suitable for the interpretation of global and local phenomena. Real-time path finding and contouring tools are the first tool sets to have been prototyped in the Centre.
For details of ongoing work go to Tessa's PhD project page.