ILAPF: Incremental Learning Assisted Particle Filtering

TL;DR

ILAPF introduces an incremental learning approach to particle filtering that adaptively estimates outlier ranges, enhancing accuracy and speed in dynamic system state estimation with noisy measurements.

Contribution

The paper presents a novel ILAPF method that learns outlier ranges incrementally, improving robustness and efficiency over existing robust particle filters.

Findings

ILAPF outperforms recent robust particle filters in accuracy.

ILAPF is faster in convergence.

The incremental learning property enables transfer learning among related tasks.

Abstract

This paper is concerned with dynamic system state estimation based on a series of noisy measurement with the presence of outliers. An incremental learning assisted particle filtering (ILAPF) method is presented, which can learn the value range of outliers incrementally during the process of particle filtering. The learned range of outliers is then used to improve subsequent filtering of the future state. Convergence of the outlier range estimation procedure is indicated by extensive empirical simulations using a set of differing outlier distribution models. The validity of the ILAPF algorithm is evaluated by illustrative simulations, and the result shows that ILAPF is more accurate and faster than a recently published state-ofthe-art robust particle filter. It also shows that the incremental learning property of the ILAPF algorithm provides an efficient way to implement transfer learning among related state filtering tasks.