“…In the literature, the robust smoothing problems mainly consider two situations. In the first case, the noise distribution is known but it is not necessarily Gaussian [29,30,31,32,33,34,35,36,37,38], for instance the noise process is assumed to have a non-Gaussian distribution in order to model outliers, temporary model uncertainties, missing observations or sensor delays. Some of these robust paradigms are adaptive because the parameters of the noises characterizing the state space model are inferred from the collected data.…”