Autonomous navigation and environment modeling for MAVs in 3-D enclosed industrial environments

摘要

In many applications, the industrial environments are typically 3-D indoor spaces enclosed by shell style structures, which are highly complex with known or unknown non-convex obstacles. GPS signal is unreliable or even unavailable inside, which poses significant technical challenges for the state estimation of micro aerial vehicles (MAVs) performing exploration and modeling tasks in such environments. In this paper, requirements and challenges for 3-D enclosed industrial environments exploration are analyzed firstly, and then state-of-art developments of MAV systems, environment modeling, visual navigation and guidance technologies are reviewed. A robust RGB-D odometry is introduced into the system to provide airborne 6-DOF state estimates of the MAV, which are fused with inertial measurements. Then the fused state information is used to assist the RGB-D based real time 3-D environment modeling. An improved closed-loop RRT based path planning approach (Bl-RRT) is developed for information-efficient environment explorations. A flight experimental platform is constructed and the proposed system is validated in flight experiments.