An agent architecture for an unattended ground sensor network

The technical evolution the last decade with the outcome of new small sensors has enabled the development of unattended ground sensor networks. The missing part is an architecture that enables autonomous detection, tracking and classification of targets in the covered area while minimizing communication and sensor usage. A sensor network architecture, with mobile track agents that moves through the network as the tracked target moves, has been developed and implemented. Dispatching of sensor data, fusion and power management are also handled by the agents. Simulated sensors produce the sensor data from targets in the simulation environment. To support different sensor types to be connected, a general abstraction level of uncertainty areas and sensor models has been introduced. The result is a general, robust sensor network system that can detect and track multiple targets and scales well when increasing the number of sensors and computational nodes in the network.