Multiple Model Kalman and Particle Filters and Applications: A Survey

Abstract

Abstract Kalman Filters (KF) is a recursive estimation algorithm, a special case of Bayesian estimators under Gaussian, linear and quadratic conditions. For non-linear systems, Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) provide first and higher order linearization approximations. Particle Filters (PF), on the other hand, are sequential Monte Carlo methods to provide estimations for non-linear non-Gaussian problems. For complex systems, Kalman or Particle Filter based single model filters may not be sufficient to model the system behaviour. Multiple Model (MM) Filters achieve more reliable estimates by using more than one filter with different models in parallel and the outputs of each filter are fused by assigning a probability to each filter. The most common methods used in the literature for multiple model estimation are Multiple Model Adaptive Estimation (MMAE) and Interacting Multiple Model (IMM). This paper presents an overview of the recent research on multiple model filters.