Research on the adaptive mechanisms in particle filter

Abstract

Particle filter based on the adaptive mechanisms has become a key issue for the recursive Bayesian estimation problem with non-linear, non-Gaussian and multi-modal distribution. Aimed at the inherent deficiency in particle filter and combined with the up-to-date research and application in the mobile robot field, some key technologies in current study are respectively summarized from the adaptive mechanism of sample size, the resampling strategy, proposal distribution, motion / likelihood model and the integration with other methods. At the same time, the main challenges that need to be solved in this field are concluded and some future trends about the technology of these difficulties are also presented.