"Prediction using Particle Filters and HMM models with correlated measure and dynamic noises”

PEDRO GADZE; BRUNO CERNUSCHI FRÍAS

Rosario, Argentina

Congreso; XIII Reunión de Trabajo en Procesamiento de la Información y Control, RPIC 2009, 16 al 18 de Septiembre de 2009, Rosario, Argentina, ISBN 950-665-340-2, pp. 547–552.; 2009

Universidad Nacional de Rosario

Sequential simulation-based methodsfor Bayesian filtering of non-linear andnon-Gaussian dynamic state-space models involvesrecursive estimation of filtering and predictivedistributions of unobserved time varyingstate process based on noisy observations.This work introduces a new model where themeasurement process and the next state areconditionally dependent given the present stateand in it we look for the predictive distributionof the state. The required distribution of thestate is represented as a set of random samples.A simulation example of a model with dynamicand process noises jointly Gaussian distributedis presented. For this case, the performanceof the proposed algorithm is very close to theperformance of the optimal filter, which is theKalman filter for this case.