Optimization Based Constrained Gaussian Sum Unscented Kalman Filter∗

This work presents a novel constrained nonlinear state estimation approach for nonlinear dynamical systems. The proposed approach combines two key elements from well know Gaussian Sum Unscented Kalman Filter (GS-UKF) and Unscented Recursive Nonlinear Dynamic Data Reconciliation (URNDDR) approaches. The proposed approach uses sum of Gaussians representation in GS-UKF and explicit constrained update in URNDDR to obtain feasible state estimates. The benefits of the proposed approach are demonstrated over the available constrained GS-UKF variants using a three state isothermal batch process case study available in literature.