Proceedings-Artikel
On Accelerating the ss-Kalman Filter for High-Performance Computation
C. PĂ©rez; L. Gracia; N. GarcĂa; J.M. Sabater; J.M. AzorĂn; JosĂ© de Gea Fernández
In: J.M. Corchado; S. RodrĂguez; J. Llinas; J.M. Molina (Hrsg.). International Symposium on Distributed Computing and Artificial Intelligence. International Symposium on Distributed Computing and Artificial Intelligence (DCAI-2008), October 22-24, Salamanca, Spain, Pages 132-141, Advances of Soft Computing, Vol. 50/2009, ISBN 978-3-540-85862-1, Springer, Berlin/ Heidelberg, 9/2008.
Abstract
This paper presents the equations of the steady state Kalman Filter (ssKF) for both variable and constant sampling times, in order to state how important it is for the stability of this lter to have a constant sampling time. Under the condition of a constant sampling time (achieved here by using recon gurable hardware), the steady-state Kalman Filter is then rewritten using a matrix property that will allow an efficient implementation in a parallel processor (although not in a sequential one), substantially improving the lter performance. This work also presents the solution to the particular cases for the propagation of the lter which can be found when implementing the algorithm, and demonstrates that the error introduced by using a xed-point numerical implementation is stable with time.