Nov 22, 2016, 4pm
Jonathan Jonker, Department of Mathematics, University of Washington
Abstract: The Kalman filter is a process of estimating unknown variables using a sequence of known measurements. It is widely used in fields such as navigation, tracking and economics. While this is not classically an optimization problem, it has been shown that by using a maximum a posteriori estimate it can be reformulated to be a least squares minimization problem. This reformulation assumes that certain covariance matrices are nonsingular; we show that this assumption may be dropped. Once the optimization problem is derived, we discuss solution methods and provide numerical examples.
Joint work with Sasha Aravkin.