ABSTRACT
This chapter discusses a flexible self-organizing approach to sensor network design, implementation, and tasking. Track information is used to estimate the future course of the entities and allocate sensors. Target to track matching suffered in the extended Kalman filter most likely as a result of model fidelity. The Bayesian net track formation algorithm performed adequately and comparably to the pheromone-tracking model. Pheromone tracking was able to successfully construct both tracks additionally; different pheromone trails proved to be a powerful device in differentiating between vehicle tracks. Track disambiguation is performed by evaluating the belief network for every combination of near and far tracks. The track information is ambiguous when the vehicles deposit identical pheromones. The ColTraNe is a fully distributed target-tracking system. It is a prototype implementation of the theoretical intercluster-distributed tracker presented earlier. An angle gate, which automatically excludes continuations of tracks when velocity estimates show targets are moving in radically different directions, has been inserted into the track-matching metric.
