Tightly coupled Visual Inertial Odometry based on Artificial Landmarks

Motion planning and control of mobile robots rely on high-accuracy estimation of pose and velocity. Many researchers use motion capture system to estimate the robot state, but this system is usually expensive and can only be used indoor. For this reason, this paper proposes an artificial landmarks based cost-effective visual inertial odometry system where Iterative Extended Kalman Filter (IEKF) acts as the back-end optimizer. Since most tags used as artificial landmarks like bar code facing the problem of decoding the information during the detection, we redesign the tag which can simplify the deployment and improve the detection efficiency. The proposed IEKF filter framework makes the measurement of the fisheye monocular camera and the IMU combined to co-estimate the pose of robot, the location of detected corners in tag and the IMU bias. Additionally, on-line calibration of the IMU bias and the extrinsic parameters between IMU and camera can be done during the robot motion. By running our algorithm real-time on unmanned aerial vehicle (UAV) and comparing with the groundtruth data obtained from the Optitrack motion capture system, our algorithm can provide high-precision estimation of the pose and velocity of UAV.