Towards Fully Autonomous Visual Inspection of Dark Featureless Dam Penstocks Using MAVs
Published in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016
This paper presents a visual-inertial navigation system for fully autonomous inspection of dark, featureless dam penstocks using a custom-built hex-rotor MAV. The system fuses data from four onboard cameras and two lidars to estimate 6-DOF pose in GPS-denied, low-visibility environments, with one camera dedicated to tracking along the tunnel axis. The multi-sensor fusion is performed via an Unscented Kalman Filter, enabling stable navigation in symmetric, textureless geometries. The approach is validated through field deployments at Carters Dam (GA) and Glen Canyon Dam (AZ), demonstrating autonomous flight, panoramic imaging, and pose estimation with ground truth comparison.
Recommended citation: Özaslan, T., Mohta, K., Keller, J., Mulgaonkar, Y., Taylor, C. J., Kumar, V., Wozencraft, J. M., & Hood, T. (2016). "Towards Fully Autonomous Visual Inspection of Dark Featureless Dam Penstocks Using MAVs." In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4998–5005.
Download Paper
Related people: Tolga Ozaslan