A batch filter and state estimator for poorly modeled distributed parameter systems

In this paper, a new filter and state estimator is developed for distributed parameter (DP) systems with parametric and structural model uncertainty. Distinguishing features of the new filter include its ability to account for realistic modeling errors, and its robust behavior in the presence of significant measurement noise. The case of spatially discrete observations, continuous in time is presented for both linear and nonlinear system models. The results are easily extended for use with other sampling methods. By way of example, the filter is shown to significantly outperform the classical DP Kalman filter.