Monocular Visual-Inertial State Estimation for Micro Aerial Vehicles

Abstract

Autonomous micro aerial vehicles (MAVs) equipped with onboard sensors, are idea platforms for missions in complex and confined environments for its low cost, small size and agile maneuver. Due to the size, power, weight and computation constraints inherent in the filed of MAVs, monocular visual-inertial system that consist of one camera and an inertial measurement (IMU) are the most suitable sensor suit for MAVs. In this paper, we proposed a monocular visual-inertial algorithm for estimating the state of a MAV. Firstly, the Semi-Direct Visual Odometry (SVO) algorithm used as the vision front-end of our framework was modified so that it can be used for forward-looking camera case. Second, an Error-state Kalman Filter was designed so that it can fuse the output of the SVO and IMU data to estimate the full state of the MAVs. We evaluated the proposed method with EuRoc Dataset and compare the results to the state-of-the-art visual-inertial algorithm, VINS-Mono. Experiments show that our estimator can achieve comparable accurate results.