A Robust and Efficient EKF-based GNSS-Visual-Inertial Odometry

Abstract

Reliable outdoor navigation is a critical technology in a wide range of applications such as autonomous driving and unmanned vehicles. Low-cost GNSS-Visual-Inertial-Odometry (GVIO) systems have received great attention from researchers since that they can achieve accurate global state estimation without drift. Nonetheless, The performance of the current algorithm is not good enough in the scene with severe GNSS occlusion, and the computational efficiency needs to be improved. In this paper, we present an EKF-based framework to tightly couple visual images, GNSS raw observation and inertial measurements. We conduct extensive experiments on various scenarios including open areas and complex indoor-outdoor switching environments, whose results have demonstrated that our method outperform existing GVIO systems in terms of localization accuracy and computation efficiency.