Kalman Filter, Unscented Filter and Particle Flow Filter on Non-linear Models

TL;DR

This paper derives and implements Kalman, Unscented Kalman, and Particle Flow Filters for non-linear models, comparing their effectiveness to guide suitable applications in stochastic variable prediction.

Contribution

It provides a detailed derivation and implementation of three advanced filters and compares their performance for non-linear stochastic models.

Findings

Kalman Filter performs well in linear models.

Unscented Kalman Filter handles non-linearities better.

Particle Flow Filter offers advantages in complex models.

Abstract

Filters, especially wide range of Kalman Filters have shown their impacts on predicting variables of stochastic models with higher accuracy then traditional statistic methods. Updating mean and covariance each time makes Bayesian inferences more meaningful. In this paper, we mainly focused on the derivation and implementation of three powerful filters: Kalman Filter, Unscented Kalman Filter and Particle Flow Filter. Comparison for these different type of filters could make us more clear about the suitable applications for different circumstances.