Proposta de um filtro de partículas aliado ao filtro de Kalman estendido iterativo para estimação de estados de sistemas não lineares com ruído Gaussiano

Abstract

About the century of 1900, control systems techniques using states feedback began to get on the highlights. Such techniques need the state vector to be avaliable, what is not always possible to do with measurement equipments. So, techniques which implement state estimation became the center of attention of researchers. These state estimators use the system dynamic information and the input and output signals to estimate the states. The state estimator known as Kalman filter is the most acceptable and useful solution to linear systems and it is acknowledged as the solution to linear systems state estimation problem. Nonlinear systems, however, have no generic estimation method defined. The most famous nonlinear technique has been the extended Kalman filter, which is the first choice of application to many systems. On 1990s, another technique called particle filter got the spotlights, because the technological improvement allowed its implementation. The particle filter has been a technique which has shown good results on nonlinear systems state estimation. In this dissertation, it is proposed a particle filter with sampling importance resampling allied to iterated extended Kalman filter (FPA-FKEI) to nonlinear systems state estimation. The efficiency of the proposed method is proven through Monte Carlo realizations in 3 systems, a monovariable, a inverted-pendulum car and an electrical power system with 4 generators.