Robust Gaussian particle filter based on modified likelihood function
Kailong Li, Lubin Chang · IET Science Measurement & Technology · 2017
This study develops a robust Gaussian particle filter (RGPF) based on modifying the likelihood function by Huber's M‐estimation theory. In the developed RGPF, the innovations are reweighted based on the Huber's cost function, resulting in a modified likelihood function which is then used to update the weights of the involved particles. In the normal case, the developed RGPF has a comparable performance in terms of accuracy with the original GPF and better filtering consistency. When there are outliers and contaminated distributions in the measurements, the RGPF can outperform the GPF in terms of both accuracy and consistency. The validity of the developed algorithm is demonstrated through numerical simulation studies.