Smart Embedded Network of Sensing Entities

Workpackage 1 encapsulates the application context. This workpackage concerns itself with developing an in-depth understanding of the context, what information should be forwarded to system operators and how, which concerns and problems this specific application offers, and how to formulate tests and evaluations in the application context.

Workpackage 2 specifies the overall architecture of SENSE. This workpackage also develops the tools required to understand system behavior and evaluate its performance. Such tools are required because the distributed nature of knowledge in the system and the way in which it is developed through local sharing make it difficult to understand the system behavior simply by observing high-level behavior.

Workpackage 3 is the development of the embedded system platform for the test bed. The platform incorporates both acoustic and visual sensors together with the necessary hardware and software processing capabilities. The acoustic sensors are microphones (or arrays to give the system access to additional spatial information). The intelligence processing mechanisms in the system (workpackages 3, 4 and 5) is embedded in the hardware platform.

Workpackage 4 is concerned with feature extraction. It builds on existing algorithms to identify and track audio and visual objects. Work on the two modalities runs in parallel and mirrors each other in order to provide conceptually consistent information to the higher level processing.

Workpackage 5 The high-level categorization of events and the learning/generalization of their behavior is undertaken. The workpackage centre on the use of rule-based and probabilistic models of objects and their behavior, built using both a priori knowledge and learning, and data from a single device's sensory environment. Issues addressed include: learning the environment of a sensor,event representation, building probabilistic models of object behaviors, and local model updating.

Workpackage 6 centers on the exchange of information between the distributed devices. Issues addressed include:

• inter-sensor communication,
• deriving a consistent overall view of the environment from local sensor information,
• identifying the equality of events and objects that different sensors and devices may perceive with radically different characteristics,
• updating the internal view of each sensor device so that they adapt to reflect changes in the SENSE system and it environment,
• and the propagation of events and other information through the distributed system to a user or supervisor.

Workpackage 7 comprises the system integration. As system components become available within the workplan they are integrated and tested to ensure the functionality and interoperability. Integration is undertaken primarily bottom-up throughout the project but provision has been made in workpackage 7 for the collection of initial test data prior to the availability of the hardware test platform. This enables design ideas to be tested ahead of full commitment to hardware platform specification.

Workpackage 8 covers the system validation and experimentation. This work assumes far more importance in the project than it would have done if a traditional systems engineering approach had been used to the development the system. In a traditional system the overall functionality can be predicted from the design and performance of the component parts. Whereas for the SENSE system it is expected that insights are gained from observing its behavior during actual testing of the system.