Collision avoidance system with situational awareness capabilities for autonomous MAV indoor flights

Micro Aerial Vehicles (MAVs) lend themselves to perform supporting reconnaissance flights and provide rescue forces with important information beyond their own field of vision. In this paper, we present an auto-adaptive collision avoidance system to aid indoor reconnaissance missions of autonomous MAVs by means of situational awareness. Efficient multi-sensor indoor classification and landmark detection on the fly serve an Artificial Potential Field (APF) based motion control. In applying supplemental potentials tailored to the specific needs of the detected surrounding situations, control factor calculation remains merely reactive. Yet, the resulting virtual force field will imply simultaneous obstacle avoidance and intelligent flight control. The system is depending only on data from camera, 2D laser rangefinder and inertial measurement unit. Neither a data link to the ground station nor any training data is required to independently explore accessible interior space in unknown environment. Finally, availability for use is validated by both simulated and real flight data.