The marginalized particle filter is a powerful combination of the particle filter and the Kalman filter, which can be used when the underlying model contains a linear substructure subject to Gaussian noise. This paper surveys state of the art for theory and practice.