Filtro de partículas baseado em estimação adaptativa inovativa

Abstract

In this work, a proposal of a particle filter based on innovative adaptive estimation for nonlinear systems is presented, as well as its fundamentals and operation. With the objective of mitigating the uncertainty or the lack of knowledge of the covariance matrices of process and measurement noises, simultaneously, the particle filter is allied to the innovative adaptive estimation. For such purpose, the difference between the theoretical and measured innovation covariances is defined as an approximation that uses the average of a moving estimation window for the innovation sequence calculus. This difference is computed continuously, using innovative adaptive estimation based on maximum likelihood theory to dynamically adjust the covariances of the particle filter. To illustrate the efficiency and applicability of the proposed filter, simulations are carried out for estimating the state of a system considering different scenarios, with and without uncertainty. The simulations results show that the proposed filter performs well in terms of robustness compared to extended Kalman filter and classic particle filter when uncertainty about process and measurement noises increases.