Identification of Linear State-Space Models Subject to Truncated Gaussian Disturbances

Within Bayesian state estimation, an important effort has been put to incorporate constraints into state estimation for process optimization, state monitoring, fault detection and control. Nonetheless, in the domain of state-space system identification, the prevalent practice entails constructing models under Gaussian noise assumptions, which suffer from inaccuracies when the noise follows bounded distributions. This poster introduces a novel data-driven method rooted in maximum likelihood principles, aimed at identifying linear state-space models subject to truncated Gaussian noise. This approach enables the concurrent estimation of model parameters, noise statistics, and noise truncation bounds, by solving a series of quadratic programs and nonlinear sets of equations.