Autonomous Physical Inspection of Exhaust Shafts and Smokestacks Using a Fully-Actuated UAV

Multirotor UAVs are being widely employed for periodic inspections of large-scale industrial facilities to protect workers from hazardous conditions such as precarious heights and toxic chemicals. The government apparatus that ushered in the era of nuclear weapons and energy combines and compounds those hazards into enormous buildings and very high-consequence materials, many now reaching end-of-life. The decontamination and decommissioning of such structures sometimes requires inspection for low-levels of high- and low-energy radioactive materials and other toxic components. This often requires physical interaction for sampling or cleaning operations. However, manual control of physical interaction with UAVs requires experienced operators, induces fatigue, and the radio communication is often impeded by thick, reinforced walls that shield the materials. In this paper, an autonomous physical interaction control strategy to interact with the inside walls of exhaust shafts using fully-actuated UAV is proposed. In this method, the physical interaction is achieved by approaching normal to the surface, which is a precursor to establish a steady contact using hybrid physical interaction control. Estimating the local surface normal using the indoor positioning systems is limited by the low illumination and geometrical symmetry. Therefore, we estimate the local surface normal using the on-board 2D lidar and autonomously control the yaw orientation, along with planar position, in the approach phase. For mechanical simplicity, a rigid arm is attached to the fully-actuated UAV to perform contact-intensive tasks on the inside walls. Experiments were carried in laboratory and mock-up shaft of Dept. of Energy, to show the effectiveness in performing a contact-intensive task inside shafts.