N??nlinear distributed parameter estimati??n

The problem of estimating states in noisy nonsecond order distributed parameter systems, with non-linear observation, is considered for disturbances in the volume only. A least squares estimation criterion is assumed and the resulting boundary value problem is embedded in a more general class of problems. The filtering problem is solved using running stage invariant embedding, and the fixed point and fixed lag smoothing using fixed stage invariant embedding. Fixed interval smoothing algorithm is developed using a backward sweep. It is found that the smoothing estimates are independent of their error covariances. Smoothing error covariances are computed approximately using a linearization procedure. The solutions so developed degenerate into the exact solution for distributed parameter estimation if linear models are considered.