A Comparison of Nonlinear Estimation Methods for Tracked Vehicle with Slipping

Four different nonlinear filters are used to estimate both states and time-varying slipping parameters of the created kinematic model of a tracked vehicle. The first filter is the well-known extended Kalman filter. The second filter is a recent development of unscented version of the Kalman filter. The third one is a particle filter using the unscented Kalman filter to generate the importance proposal distribution. The last one is a novel and guaranteed filter that use a linear set-membership estimator and can give an ellipsoid set in which the true state lies. The four different approaches have different complexities, behavior and advantages that are compared in simulations.