A new architecture for simultaneous localization and mapping: an application of a planetary rover

Abstract

A new architecture implements one Monocular Simultaneous Localization and Mapping (SLAM) system to track the unconstraint motion of a mobile robot. The modified ORB (Oriented FAST and Rotated BRIEF) features represent the landmarks for designing a grid feature detection algorithm. An upgraded feature matching method has improved the robustness of feature matching. The Modified coVariance Extended Kalman Filter (MVEKF) estimates the multiple dimension states of the free moving visual sensor instead of the familiar Extended Kalman Filter (EKF). The simulation navigation of Lunar and Mars surfaces proves that the proposed method is robust and efficient.