Speaker's biography: Pavel Kulmon received the Master of Engineering degree in the field of geomatics and Ph.D. degree in systems engineering from the Czech Technical University, Prague, Czech Republic.,He is a Senior Researcher with the Department of Research and Analysis, ERA. His main research interests include in the field of statistical inference with focus on bayesian methods, together with machine learning, and reinforcement learning applications.

This presentation focuses on the problem of multi-static passive sensor scheduling. VERA-NG is a super-heterodyne sensor whose key point is to traverse through the frequency spectrum so that all the user requirements are fulfilled. The three most important requirements are surveillance, target tracking, and electronic support measures. First, the VERA-NG sensor is introduced and the historical overview of its development is presented. Then we proceed to show how the problem of sensor scheduling is different from the usual radar resource management (RRM) problem. We translate requirements (or tasks) into the information-theoretic domain. Having them in this common domain, we then follow with the formulation of the sensor scheduling as the multi-criteria optimization problem which produces the optimal tuning plan for a chosen time horizon. Using simulated data, we illustrate some advantageous properties of such a tuning plan and present plans for future development. Our approach has shown to be scalable in terms of the number of sensors and can be translated into many different domains where such a super-heterodyne sensor is used.

Speaker's biography: Lennart Svensson is a Professor of Signal Processing with the Chalmers University of Technology. His main research interests include machine learning and Bayesian inference in general, and nonlinear filtering, deep learning, and tracking in particular. He has organized a massive open online course on multiple object tracking, available on edX and YouTube, and received paper awards at the International Conference on Information Fusion in 2009, 2010, 2017, and 2019.

Tracking, localization, and mapping are interrelated problems that depend on accurate object or feature detections to estimate key parameters. In this presentation, we describe how Poisson multi-Bernoulli mixtures enable us to handle all these problems in a unified and fully Bayesian manner. We will discuss how the data association hypotheses impact the posterior distributions and present a range of numerical techniques for approximating these posteriors. Our approach demonstrates how PMBMs can produce state-of-the-art estimators, advancing the fields of tracking, localization, and mapping.