A Robust Interacting Multiple Model Smoother with Heavy-Tailed Measurement Noises

Interacting multiple model (IMM) estimator has found an increasingly wide utilization in the filed of target tracking. Smoothing, uses all obtained measurements to estimate the previous state in order to provide a better estimation. However, IMM smoothers do not perform well and suffer severe performance degradation with some outliers existing in the measurement noises. This paper proposes a novel robust IMM smoother to deal with heavy-tailed measurement noises obeying the Student's t-distribution. The proposed smoother is derived from a forward filter combing with a backward filter. An example of maneuvering target tracking with heavy-tailed