An approach to combine autonomous robotic inspection
with Augmented Reality–enabled human teleoperation is
proposed and aims to address the problem of robotic inspection
of structures and environments for which prior
knowledge exists but it is relatively outdated or does not
encode areas of interest and other important features. The
method employs a previously proposed structural inspection
planner which derives the initial inspection trajectory
given any prior knowledge of the environment. Live feed
of the stereo camera frames combined with the real–time
3D reconstruction of the environment is provided to the
human operator via the Augmented Reality interface. Once
an area that needs special attention, or is not 3D–mapped and
reconstructed at sufficient levels of fidelity by the initial robot
trajectory is detected, then head motion may be employed
to adjust the path of the robot and collect the required
viewpoints. The overall framework is demonstrated with the
use of an aerial robot capable of GPS–denied navigation and
online mapping of its environment.