An Online Calibration 2-DoF Attitude and Height State Estimator

The Visual-inertial based simultaneous localization and mapping (SLAM) method has been a hot topic in recent years. When Unmanned Aerial Vehicle (UAV) detects in the unknown environment, stable control and mapping is need. It is intolerable for the error of roll and pitch estimation of the attitude, and the altitude needs to maintain continuous and effective observations to prevent drift. In this paper, based on nonlinear optimization methods with a sliding window, we propose an visual-inertial odometry that solves more precise pitch and roll by IMU using EKF algorithm and integrates it into the tightly-coupled state estimation to keep 2-DoF attitude estimation precision, while simultaneous calculating the 6-DoF state by IMU pre-integration and vision. Meanwhile, the height radar is used to eliminate the cumulative drift of the drone height. A terrain estimator is set to prevent terrain abrupt changes.