Machine Learning in Predictive Filtering.

Autor: Schoenke, Jan H., Brockmann, Werner
Předmět:
Zdroj: Proceedings. 22. Workshop Computational Intelligence; 2014, p55-74, 20p
Abstrakt: This paper considers the Simultaneous State and Parameter Estimation Problem (SSPEP) for non-linear dynamic systems and its stability issues due to a possibly harmful interaction between the two estimation task. Common approaches to state estimation in general and the SSPEP in particular are reviewed. A new approach to the SSPEP is proposed to avoid the harmful interaction in SSPEPs. Experiments with simulated data highlight properties of the proposed approach and compare the estimation performance on a SSPEP with related work. The proposed approach appears to be robust to missing knowledge for initialization. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index