This study introduces an approach for and the challenges in employing unmanned aerial vehicles (UAVs) for material handling in the emerging industrial custom manufacturing environments. Compared with conventional industrial robotic systems, UAVs offer enhanced flexibility for the design and on-the-fly variation of the pathways and workflow to optimally perform multiple tasks on demand, besides offering favorable cost and dimensional footprint factors. A fundamental challenge to the deployment of UAVs in manufacturing and other indoor industrial settings lies in ensuring the accuracy of a drone's localization and flight path. Earlier approaches based on using multiple sensors (e.g., GPS, IMU) to improve the localization accuracy of UAVs are considered ineffective in indoor environments. In fact, few investigations have tackled the issues arising due to the limited space and complicated components and moving entities, human presence in shop-floor environments. Towards addressing this challenge, a pose estimation method that employs just a single camera onboard with a UAV, together with multiple ArUco markers positioned strategically over the shop-floor is implemented to track the real-time location of a UAV. A Kalman filter is applied to mitigate noise effects for pose estimation. To assess the performance of this method, several experiments were carried out in Texas A&M University's manufacturing labs. The result suggests that Kalman filter can reduce the variance of pose estimation by 88.48% compared to a conventional camera and marker-based motion tracking method (~ 27 cm) and can localize (via averaging) the position to within 8 cm of the actual target location.
- Bukkapatnam, Satish Professor