Review of Kalman filter variants for SLAM in mobile robotics with linearization and covariance initialization

Abstract

Simultaneous localization and mapping (SLAM) has become a foundational concept in robotics navigation which enabling autonomous systems to build maps of unknown environments while estimating their own position. This article aims to provide a comprehensive review of the SLAM concept in the context of mobile robotics navigation by focusing on theoretical principles, estimation problems, algorithms involved, and related applications. The existing literature is systematically analyzed and classified based on three main perspectives of navigation, which are localization, mapping, and path planning. Particular attention is given to Kalman filters and their variants in SLAM-based systems, along with crucial consideration of the linearization and covariance initialization. This article identifies the strengths and limitations of current SLAM approaches. Therefore, this article concludes by outlining research gaps and recommending directions for future exploration, with the intention of serving as a foundation for continued innovation in SLAM-based robotic navigation systems.