Перспективи використання інваріантного розширеного фільтру Калмана для автономної навігації за допомогою візуально-інерційного SLAM

Abstract

This work is devoted to the study of the application of the invariant extended Kalman filter (IEKF) in visualinertial SLAM systems for autonomous navigation. A comparative analysis of the traditional extended Kalman filter (EKF) and its modification IEKF is carried out to assess their effectiveness in localization and mapping tasks. Particular attention is paid to the ability of IEKF to take into account system symmetries to increase the accuracy of position and orientation estimation, as well as to reduce drift. The results of the study demonstrate that IEKF outperforms EKF, especially in conditions of nonlinearities and irreversible state changes, providing better model consistency and minimizing error accumulation. It is shown that IEKF-SLAM is a promising solution for autonomous vehicles, mobile robots, and other applications that require reliable navigation in complex environments. The key direction of further research is the optimization of the computational costs of IEKF and its adaptation for use in embedded systems with limited resources.