Length: 3 hours
Intended Audience: Students and researchers interested in the topic of state estimation and control in distributed and networked systems with limited hardware resources.
Description: How can we achieve high performance on embedded platforms with limited computing and communication resources? This question represents a key challenge in intelligent systems research, for example, when high-performance control must run on low-power computing hardware, or multiple agents share a wireless communication network. Traditional periodic sampling methods are inherently limited: data is processed or transmitted at a-priori fixed time instants, irrespective of whether there is any need for an update or control action, or not. The event-based sampling paradigm, which has received a lot of attention recently in controls and signal processing, addresses this limitation by performing computation and communication only when necessary as indicated by system-inherent events (for example, an error passing a threshold level, or estimation uncertainty growing too large). With event-based methods, average usage of resources can significantly be reduced compared to traditional periodic designs. Hence, event-based methods allow the designer to achieve high system performance with reduced resource usage.
This tutorial will provide an introduction to the problem of event-based state estimation (and control). In particular, we shall consider a scenario where multiple distributed agents observe a dynamic process and share sensory data over a network in order to solve a joint state estimation and sensor fusion problem. We review recent developments in the area and highlight important theoretical and technical challenges. In particular, we discuss the key aspects of event-based system design: (i) the design of event-triggering mechanisms, (ii) (sub-)
optimal estimation and filtering algorithms, and (iii) distributed architectures. In addition to
developing important problems and theoretical results, we shall also highlight different
successful experimental applications, for example, in networked control and robotics.
Prerequisites: Basic knowledge in mathematics (probability, linear algebra), dynamic systems (linear systems),
and state estimation and filtering (e.g., Kalman filtering).
Presenter: Sebastian Trimpe
Sebastian Trimpe is a Research Scientist and Group Leader at the Max Planck Institute for
Intelligent Systems in Tübingen, Germany. He obtained his Ph.D. (Dr. sc.) degree in 2013 from
ETH Zurich under the supervision of Raffaello D’Andrea at the Institute for Dynamic Systems
and Control. Before, he received a B.Sc. degree in General Engineering Science in 2005, a M.Sc.
degree (Dipl.-Ing.) in Electrical Engineering in 2007, and an MBA degree in Technology
Management in 2007, all from Hamburg University of Technology. In 2007, he was a research
scholar at the University of California at Berkeley. Sebastian is recipient of the General
Engineering Award for the best undergraduate degree (2005), a scholarship from the German
Academic Scholarship Foundation (2002-2007), the triennial IFAC World Congress Interactive
Paper Prize (2011), and the Klaus Tschira Award for achievements in public understanding of
science (2014). Sebastian has taught a graduate-level class on Recursive Estimation at ETH
Zurich and given various presentations on the topic of distributed and event-based state
estimation. See his website http://trimpe.is.tuebingen.mpg.de for more information on his
research, educational, and outreach activities.