OPTIMAL MODEL ESTIMATION DEVICE, METHOD AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To enable estimation of an optimal state space model with a nonlinear model as a target.SOLUTION: A state estimation unit 22 sequentially estimates a state vector X on the basis of time-series data of an observation value and calculates prediction errors, in accordance with a Normalized UFK (Unscented Kalman Filter) using an estimated error variance-covariance matrix, a covariance matrix of system noise and noise variance of observation noise each normalized with diagonal elements of the covariance matrix of the system noise and the total sum of variance values of the observation noise. An optimal model estimation unit 23 estimates an optimal state space model on the basis of each prediction error calculated by the state estimation unit 22 when each combination of the normalized system noise covariance matrix and observation noise variance is used.