Overview of Filtering Algorithms for Autonomous Mobile Robots

Abstract

In this paper, three filtering algorithms in SLAM for autonomous mobile robots will be referred to, which are Kalman Filter Family (Kalman Filter and Extended Kalman Fiter), Particle Filtering (PF) and Rao-Blackwellized Particle Filter (RBPF). For a mobile robot, localization and mapping are the key indicators that determine whether it can be called “autonomous” or not. The algorithm is the most important of them all. In the case of SLAM algorithms, there exist three algorithms based on optimization, based on filtering, and based on Georgia Tech Smoothing and Mapping (GTSAM). The filtering algorithm, as the oldest algorithm, is the focus of this paper. And a large number of existing studies in the broader literature have examined that Bayesian filtering is the basis of KF, EKF and PF. In the following paragraphs, in order to better represent the development process of filtering algorithms, we will start with Bayesian filtering and introduce its theoretical framework, as well as the origin and derivation process of several subsequent filters.