Current status and future challenges of moving horizon state estimation

March 01, 2013, Webb 1100

James Rawlings

University of Wisconsin-Madison, Department of Chemical and Biological Engineering


This talk presents the fundamentals of moving horizon state estimation (MHE).  A brief historical overview of on-line optimization as an enabling technology for state estimation for nonlinear dynamical systems is provided.  The current state of the theory for MHE is summarized. We focus on the property of robust global symptotic stability (RGAS) for state estimation and establish a new result using this definition, namely that full information estimation is RGAS for the case of a nonlinear detectable system subject to convergent state and measurement disturbances. The advantages and disadvantages of MHE compared to popular alternatives such as the extended Kalman filter are discussed.  Some nonlinear examples are presented to display some basic tradeoffs between computational intensity and quality of state estimation.  Two unsolved research problems are presented: (i) suboptimal moving horizon estimation and (ii) establishing RGAS for the case of bounded rather than convergent disturbances.

Speaker's Bio

A longtime proponent of technology–enhanced learning, Paul A. Elfers Professor of Chemical and Biological Engineering James B. Rawlings draws on a suite of new and existing technological tools to engage students in such subjects as chemical process modeling and computational modeling of reactive systems.
In four chemical and biological engineering courses, Rawlings capitalizes on the powerful campus wireless network to transform simple
campus classrooms into interactive teaching laboratories in which he and his students use laptop computers to tackle web–based problems in real time. Among these courses is CBE 255, Introduction to Chemical Process Modeling, a required course for sophomores that launched in 2007. Rawlings collaboratively developed the course with colleagues in engineering physics and civil and environmental engineering under an engineering problem–solving with computers linked–courses project. 
        While his multifaceted teaching approach has enriched and improved engineering students’ learning experiences on campus, Rawlings
co–authored a textbook and supervised creation of a software modeling language that have benefited students and researchers around the world.