Non-centralized target tracking in networks of directional sensors

An important class of sensors is the class of directional sensors. These sensors detect targets within a range and a predefined direction. In this work, we assume that directional sensors are deployed in a sensor field, where at each node, there are four collocated directional sensors providing a coverage of 360°. The nodes of the sensors broadcast only binary information, that is, they transmit a signal only if one of their sensors detects a target. When they do not broadcast, it is assumed that their sensors have not detected a target. Each node can process the measurements of its own sensors and the information received from other nodes. We study two setups, where at anytime, a target in the sensor field is being tracked by all the nodes or only by one of them at a time. For the latter case, we propose a scheme for making a decision of passing the task of tracking to other nodes in the network. When a processing node is relinquishing processing, it broadcasts relevant information for continued tracking. From this information, other nodes unambiguously decide the next processing node. We explore the implementation of tracking by particle filtering. We demonstrate the performance of the proposed method with simulations of various scenarios.