Abstract
The seat track inspection is a laborious, time-consuming process that poses a significant bottleneck to the heavy maintenance check for aircraft. It presents a significant opportunity and hangar cost to maintenance, repair, and overhaul (MRO) companies, as well as airlines. Amidst a looming manpower crisis in the sector, and many operational inefficiencies with the current strenuous manual inspection process, there exists a unique opportunity for automation. To this end, MIRA, a first-of-its-kind robotic solution for seat track inspection was developed and validated with user feedback from the industry. Furthermore, the methyl-opencv-suite (MOS), a new rapid development suite for computer vision pipelines was created to aid in the development of Crown Counting Visual Odometry (CCVO), a novel visual odometry method specifically for use in the seat track inspection. CCVO odometry was integrated with the Robot Operating System (ROS) and physical experiments were run on a 70cm stretch of corroded aircraft seat track to characterise the performance of CCVO, and it was found to perform almost as well as an encoder under the expected operating conditions for MIRA (0.551% vs 0.670% drift over distance), showing high viability for use in industrial settings. Failure modes for CCVO were also identified, as well as potential future improvements to the algorithm.