Robust Planar Tracking via a Virtual Measurement Approach

In this paper the classical problem of planar tracking is studied. The approach here proposed is based on the idea of considering, as system output, a vector of "virtual" measurements directly obtained from the actual ones. In this way, the measurement map is split into the sum of a linear time-varying transformation of the state and an uncorrelated white noise process, which is generally nongaussian. The resulting model is amenable for applying standard linear and polynomial Kalman like algorithms avoiding any linearization procedure of the measurement map, which is required by other standard suboptimal solutions (e.g. EKF). Finally, the proposed algorithms are checked through numerical simulations.