Visual Perception With Servoing for Assisting Micro Aerial Vehicle Through a Staircase

This article presents a visual perception system with servoing design for a micro aerial vehicle (MAV) flying through a staircase. Since space in an indoor environment is usually limited, the size of MAV employed here have to be small and the payload of MAV is limited. Two apinhole cameras mounted on the front and bottom of the MAV are utilized to take turns in observing the stairs. The captured images and the flying control commands are transmitted between a remote computational processing platform and the MAV through a wireless network. Visual features of the stairs, wall, railing, and the landing are described to estimate the location and orientation of the MAV in the staircase. This visual information refined through the probabilistic data association filter is then used by the MAV as a basis for navigating its environment. A system of visual servoing based on fuzzy logic is proposed to control the MAV's flight through a set of straight stairs and align the entrance of the stairs. The MAV's ability to fly online through the staircase is verified in several experiments to demonstrate the efficiency of the overall system.