Sensor Management Using Relevance Feedback Learning
Chris Kreucher, Keith D. Kastella · 2003
An approach that is common in the machine learning literature, known as relevance feedback learning, is applied to provide a method for managing agile sensors. In the context of a machine learning application such as image retrieval, relevance feedback proceeds as follows. The user has a goal image in mind that is to be retrieved from a database of images (i.e., learned by the system). The system computes an image or set of images to display (the query). Oftentimes, the decision as to which images to display is done using divergence metrics such as the KullbackLeibler (KL) divergence. The user then indicates the relevance of each image to his goal image and the system updates its estimates (typically a probability mass function on the database of images). The procedure repeats until the desired image is found. Our method for managing agile sensors proceeds in an analogous manner. The goal of the system is to learn the number and states of a group of moving targets occupying a surveillance region. The system computes a sensing action to take (the query), based on a divergence measure called the Renyi divergence. A measurement is made, providing relevance feedback and the system updates its probability density on the number and states of the targets. This procedure repeats at each time where a sensor is available for use. It is shown using simulated measurements on real recorded target trajectories that this method of sensor management yields a ten fold gain in sensor efficiency when compared to periodic scanning. EDICS Category: 2-INFO