Using Absolute Non-Gaussian Non-White Observations in Gaussian SLAM

In the navigation context it is typical the presence of sensors that introduce uncertainties that cannot be modeled as white Gaussian noise. Such measures cannot be directly used in gaussian estimators. This paper presents a technique that allows the consistent processing of this type of measurements in combination with a standard EKF estimator. The method can be applied in an efficient implementation of SLAM based on a Gaussian estimator.