A hybrid self-adaptive particle filter through KLD-sampling and SAMCL

Abstract

The purpose of this paper is to present a hybrid method of a particle filter for localization in mobile robotics. The main references are the particle filter based on Kullback-Leibler divergence and a self-adaptive particle filter using grid-energy. Gains and drawbacks of each method are discussed and compared with the developed algorithm. This hybrid particle filter explores the best quality of each method and the final result brings a solution to the localization problem: position tracking, global localization and kidnapping in a deterministic environment. This work was developed using the ROS framework (Robot Operating System) and tested with a Pioneer 3DX robot in real and simulation environments.