1: \begin{abstract}                          

2: We propose two nonlinear Kalman smoothers

3: that rely on Student's t distributions.

4: The {\it{T-Robust smoother}} 

5: finds the maximum {\it a posteriori}

6: likelihood (MAP) solution for Gaussian process noise and 

7: Student's t observation noise, and is extremely 

8: robust against outliers, outperforming the recently

9: proposed $\ell_1$-Laplace smoother in extreme situations

10: (e.g. 50\% or more outliers). 

11: The second estimator, which we call the

12: {\it{T-Trend smoother}}, is 

13: able to follow sudden changes in the process model,

14: and is derived as a MAP solver

15: for a model with Student's t-process noise and Gaussian 

16: observation noise. We design specialized methods to 

17: solve both problems which exploit the special structure

18: of the Student's t-distribution, and provide a convergence theory.

19: Both smoothers can be implemented with only minor modifications

20: to an existing $L_2$ smoother implementation. 

21: Numerical results for linear and nonlinear models

22: illustrating both robust and fast tracking applications are presented. 

23: \end{abstract}

24: