Robust navigation of MAVs based on deeply integrated laser-camera information

Micro Air Vehicles (MAVs) are becoming very popular at present time. They are lightweight, affordable and very flexible in terms of location. Thus, MAVs are predestined for rescue missions. They are able to explore areas that are difficult to access. Nonetheless, these rescue missions are very complex. They require a reliable autonomous operation system even in GNSS-denied areas. In order to assure accurate navigation solution at all times other sensors, such as camera and laser rangefinder, are needed. In this paper, particular attention is paid to those two sensors. Hereby, a robust ego-motion estimator is established. The MAV's ego-motion is estimated by solving the Perspective-3-Point (P3P) problem. The integrated navigation system comprises keyframe architecture and uncertainty estimation. In experiments it is shown that the established navigation system can deal with very challenging environments. Beyond that, it is demonstrated that the new approach outperforms the standard laser-scan-matching approach.