报告题目：Reachability analysis in a data-driven framework
报告时间：2023年10月24日（星期二） 10:00
报告形式与地点：线下讲座 海航科技大楼D313
报告专家：Dr. Luis A. Duffaut Espinosa
讲座内容：
The abundance of information and reduction in the cost of sensors have significantly propelled technological advancements in the area of autonomy. In recent years, this has been highlighted by the big push to make control methodologies data-driven. This is indeed the case for reachability analysis, which is often used in applications to permit maintaining systems in a safe region of operation. This area has seen significant progress based on advances in set methods and dynamic programming. However, its efficiency, feasibility, and strong dependence on state space models are obstacles to expanding reachability into the data-driven scene. The problem is exacerbated in multi-agent systems applications when multiple reachable sets must be computed at once.
This talk presents a methodology for data-driven reachability analysis. Particularly, a general framework for describing the input-output behavior of systems is used in place of state space models. This is the so-called Chen-Fliess framework. For instance, any control-affine system or Volterra operator with analytic kernels can be described using this representation. This framework will be used for the efficient computation of the minimum-bounding box over approximation of the true output reachable set of dynamical systems. Several examples will be presented to illustrate the approach.
报告专家简介:
Dr. Luis A. Duffaut Espinosa is an assistant professor at the Department of Electrical and Biomedical Engineering at the University of Vermont. He received a B.S. degree in physics from the Universidad Nacional de Ingenieria, Lima, Perú, in 2003, an M.S. degree in mathematics with a mention in stochastic processes from the Pontificia Universidad Católica del Perú, Lima, Perú, in 2005, and the Ph.D. degree in electrical and computer engineering from Old Dominion University, Norfolk, VA, USA, in 2009. He held postdoctoral positions at Johns Hopkins University and the University of New South Wales (Australia) in 2010 and 2011-2013, respectively. His research lies in the intersection of control, estimation, and signal processing. One of his long goals is to develop a general framework for the control and estimation of uncertain nonlinear systems that generalize the algebra of convolutions used for linear systems. Currently, this framework serves for the systematic design of control laws, data-driven reachability analysis, and model-free Kalman filters, and the robustification of multi-agent systems applications. He is applying these methodologies to efficient real-time data assimilation and spatial reconstruction in harsh environments with limited information, model-free control of multi-agent systems. His research has been funded by the National Science Foundation, the Department of Energy, the Cold Regions Research and Engineering Laboratory, etc..
欢迎感兴趣的老师、研究生同学踊跃参会交流!
