Prof. Martin Adams (University of Chile)
Department of Statistics, LG.03
Abstract: The application of random sets in multi-target tracking has led to the development of Finite Set Statistics (FISST) which provides the basis for filters such as the Probability Hypothesis Density (PHD) filter and more recently the Generalized Labelled Multi-Bernoulli (GLMB) filter, which recently attracted considerable research interest as well as deployment in commercial applications. This presentation advocates that the same principle applies in fields as diverse as Simultaneous Localization and Mapping (SLAM) in robotics, where instead of referring to the problem of target estimation, the problem of map feature or environmental object estimation are of concern, and in space debris tracking. In SLAM, the concept of a set based measurement and map state representation, allows all of the measurement information, spatial and detection, to be incorporated into joint Bayesian and Maximum Likelihood SLAM frameworks. Representing measurements and the map state as random sets, rather than the traditionally adopted random vectors, is not merely a triviality of notation. It will be demonstrated in this presentation that a set based framework circumvents the necessity for fragile data association and map management heuristics, which are necessary, and often the cause of failure, in vector based solutions.
An example application of the field of multi-target tracking is Space Situational Awareness (SSA). This corresponds to the identification and tracking of orbiting debris travelling at dangerously high speeds, which is of concern for the safety of operational satellites and future space missions. Determining the number and state (e.g. position and velocity) of debris components which have passed through the field of view of a telescopic image, in the presence of measurement noise, clutter and false alarms, is a field of research which can be addressed with FISST based tools. Experimental results, demonstrating SLAM in urban and marine environments and also space debris tracking, based on radar images, will be demonstrated.