The edge node is the interface of the SMART system with the physical world. To do so it offers the following functionality:
- Interfaces to sensors via the sensor drivers.
- Processing of sensor streams, yielding metadata. These streams are audio-visual, but also stem from sensor and social networks.
- Fuses the metadata in an intelligent way, reasoning for events.
- Delivers information to higher layers on the basis of standard file formats.
- Can be configured on the basis of rules of automatic or manual rules.
The high-level edge node architecture is as follows:
The signal streams, either form physical sensors, or from social networks are processed by the various perceptual components, i.e. the algorithms that attempt to interpret the streams' meaning and extract context. For details on metadata extraction and the perceptual components used in SMART, see here.
The resulting metadata are fused, using initialisation information from the node setup. As a result, events of interest are extracted. These events relate for example to the dynamic behavior of crowds or traffic.
All the elements of the edge node are configurable:
- The sensors are configured via their drivers. For example, their update rate can be configured to save power based on the type or service (manual configuration) or the detected events (automatic configuration can reduce update if nothing is happening).
- The perceptual components are configured to alter their behavior on the basis of the detected conditions (environmental or event-related).
- The Intelligent Fusion Manager can be configured to adopt its reports, producing different collections.
The Intelligent Fusion Manager and its configuration is described here.
All the edge node metadata are stored in the node's database, where they are consolidated in metadata streams to be indexed by the search engine in the next layer. The handling of the metadata from and to the database is described in the edge node metadata flow.
Social Networks and Sensor Networks Integration
A social network (Twitter) has been integrated into the SMART system as a sink of information from the sensor network. This implements our vision about Things that Perceive and Tweet.
Twitter is also being used as a source of information. It is considered another sensor for SMART, and its data stream is processed in a similar manner as the rest of the perceptual components do for other sensors.
See our demo page for more information on using the social networks as information sink and source.