Searching for, initiating and tracking multiple targets using existence probabilities

We present a framework for searching for and tracking multiple targets, using existence probabilities to perform track management. Unlike other approaches, we also use existence probabilities to initiate new tracks. This is done by having a ldquosearchrdquo track which represents the probability of an unconfirmed track being present and the state distribution of such a track. The sensor can have a state-dependent probability of detecting a target, and we are particularly motivated by scenarios where a sensor periodically scans a search space, being able to see only part of the space at a time. Results show that this approach is able to confirm tracks in an environment with significant clutter levels and a low probability of detection, with a minimal number of false tracks.