I. Introduction
Visual-inertial (VI) camera pose tracking (i.e. odometry) has recently attracted significant attention in the computer vision, augmented reality, and robotics communities. The two sensor types, camera and inertial measurement unit (IMU), complement each other well in a joint optimization framework. Visual sensors provide rich information for robust visual tracking and allow for referencing the camera trajectory towards previously visited parts of the environment. Using vision, the full 3D rotation and translation of the camera can be observed if the environment provides sufficient features. IMUs, on the other hand, measure accelerations and rotational velocities at high frame-rates. This allows for overcoming degenerate visual cases such as pointing the camera at a textureless wall. IMUs also observe acceleration due to gravity and, hence, allow for an absolute horizontal reference in the environment.