Dr. Michael Zavlanos visited AME on February 2, 2017 as part of Dr. Andrea L’Afflitto’s Dream Course, Modern Control Theory and Applications.
Abstract: Current robotic systems have the potential to accomplish a previously intractable scope of tasks. Their ever growing capabilities will soon allow them to operate autonomously outside the lab, in remote, unpredictable, and uncertain environments, where the presence of humans is dangerous or even impossible. For this to become possible, a fundamental challenge is to develop new methods that will enable teams of robotic sensors to collaboratively explore unknown environments and extract concise actionable information. In this talk,we present a novel approach to dynamically synthesize optimal controllers for a robotic sensor network tasked with estimating a collection of hidden states. The key idea is to divide the hidden states into clusters and then use dynamic programming to determine optimal trajectories around each hidden state as well as how far along the local optimal trajectories the robot should travel before transitioning to estimating the next hidden state within the cluster. Then, a distributed assignment algorithm is used to dynamically allocate controllers to the robot team from the set of optimal control policies at every cluster. Compared to relevant distributed state estimation methods, our approach scales very well to large teams of mobile robots and hidden vectors. We also present a distributed state estimation method that allows mobile sensor networks to estimate a set of hidden states up to a user-specified accuracy. This is done by formulating a LMI constrained optimization problem to minimize the worst case state uncertainty, which we solve in a distributed way using a new random approximate projections method that is robust to the state disagreement errors that exist among the robots as an Information Consensus Filter (ICF) fuses the collected measurements. To our knowledge, even though the distributed active sensing literature is well-developed, the ability to control worst-case estimation uncertainty in a distributed fashion is new. We present numerical simulations and experimental results that show the efficiency of the reposed methods.
Bio: Michael M. Zavlanos received the Diploma in mechanical engineering from the National Technical University of Athens (NTUA), Athens, Greece, in 2002, and the M.S.E. and Ph.D. degrees in electrical and systems engineering from the University of Pennsylvania, Philadelphia, PA, in 2005 and 2008, respectively. From 2008 to 2009 he was a Post-Doctoral Researcher in the Department of Electrical and Systems Engineering at the University of Pennsylvania, Philadelphia. He then joined the Stevens Institute of Technology, Hoboken, NJ, as an Assistant Professor of Mechanical Engineering, where he remained until 2012. Currently, he is an assistant professor of mechanical engineering and materials science at Duke University, Durham, NC. He also holds a secondary appointment in the department of electrical and computer engineering. His research interests include a wide range of topics in the emerging discipline of networked systems, with applications in robotic, sensor, and communication networks. He is particularly interested in hybrid solution techniques, on the interface of control theory, distributed optimization, estimation, and networking. Dr. Zavlanos is a recipient of the 2014 Office of Naval Research Young Investigator Program (YIP) Award, the 2011 National Science Foundation Faculty Early Career Development (CAREER) Award, as well as Best Student Paper Awards at GlobalSIP 2014 and CDC 2006.