Certifying the optimality of a distributed state-estimation system via majorization theory
Fri., Nov 6, 2009
2:00 - 3:00 PM
380 Soda
2:00 - 3:00 PM
380 Soda
This talk will focus on the design of optimal distributed state estimators in the presence of communication costs. Our formulation comprises of two or more agents with unequal measurements of the state of a linear dynamical system driven by white Gaussian noise.
The agents can be either estimators or sensors and the overarching goal of the multi-agent system is to produce optimal state estimates in the expected squared error sense. Information flows from the sensors to the estimators via controlled erasure links. At each time step, any given sensor may choose to send information to an estimator by paying a cost or it may send a free erasure symbol. Hence, the overall cost has two components; the expected squared estimation error and a communication cost. While particular instances of this class of problems have been investigated in the literature, where suitable policies have been proposed, the challenge of proving the optimality of a given policy was open. I will discuss a method based on the theory of majorization, which allowed us to prove, for the first time, that threshold policies at the sensors and Kalman-like estimators are jointly optimal for certain decentralized state estimation problems with communication costs. Theory of majorization has been widely used in statistics and more recently in network design to prove the optimality of certain registration policies. Hence, the main purpose of this talk is to illustrate the use of majorization theory to certify the optimality of decentralized state estimation policies.
Bio:
Prof. Nuno C. Martins received the MS. degree in electrical engineering from Instituto Superior Tecnico, Lisbon, Portugal, in 1997, and the Ph.D. degree in Electrical Engineering and Computer Science from Massachusetts Institute of Technology (MIT), Cambridge, in 2004. Currently, he is Assistant Professor at the Department of Electrical and Computer Engineering, University of Maryland, College Park. He is also affiliated with the Institute for Systems Research and his research interests include fundamental limits of feedback, methods for the design of optimal networked control systems and the fusion between control theory and information theory. Prof. Martins received a National Science Foundation CAREER award in 2007 and the 2006 American Automatic Control Council O. Hugo Schuck Award (theory) and two fellowships, in 1999 and 2004, from the European Social Fund and the Portuguese Foundation for Science and Technology. He is also a member of the editorial board of Systems and Control Letters (Elsevier) and of the IEEE Control Systems Society Conference Editorial Board.
The agents can be either estimators or sensors and the overarching goal of the multi-agent system is to produce optimal state estimates in the expected squared error sense. Information flows from the sensors to the estimators via controlled erasure links. At each time step, any given sensor may choose to send information to an estimator by paying a cost or it may send a free erasure symbol. Hence, the overall cost has two components; the expected squared estimation error and a communication cost. While particular instances of this class of problems have been investigated in the literature, where suitable policies have been proposed, the challenge of proving the optimality of a given policy was open. I will discuss a method based on the theory of majorization, which allowed us to prove, for the first time, that threshold policies at the sensors and Kalman-like estimators are jointly optimal for certain decentralized state estimation problems with communication costs. Theory of majorization has been widely used in statistics and more recently in network design to prove the optimality of certain registration policies. Hence, the main purpose of this talk is to illustrate the use of majorization theory to certify the optimality of decentralized state estimation policies.
Bio:
Prof. Nuno C. Martins received the MS. degree in electrical engineering from Instituto Superior Tecnico, Lisbon, Portugal, in 1997, and the Ph.D. degree in Electrical Engineering and Computer Science from Massachusetts Institute of Technology (MIT), Cambridge, in 2004. Currently, he is Assistant Professor at the Department of Electrical and Computer Engineering, University of Maryland, College Park. He is also affiliated with the Institute for Systems Research and his research interests include fundamental limits of feedback, methods for the design of optimal networked control systems and the fusion between control theory and information theory. Prof. Martins received a National Science Foundation CAREER award in 2007 and the 2006 American Automatic Control Council O. Hugo Schuck Award (theory) and two fellowships, in 1999 and 2004, from the European Social Fund and the Portuguese Foundation for Science and Technology. He is also a member of the editorial board of Systems and Control Letters (Elsevier) and of the IEEE Control Systems Society Conference Editorial Board.
UC Berkeley Networking
Kristen Woyach and Pulkit Grover Last Modification Date: Wednesday, August 19, 2009
Kristen Woyach and Pulkit Grover Last Modification Date: Wednesday, August 19, 2009