Penstocks are long, featureless, wide, dark tunnels that carry the water from the lake to the turbines of a dam. This infrastructure requires regular and proper maintenance due to possible catastrophic consequences such as cracking of the penstock or even complete demolishment of the dam. Maintenance engineers and workers manually inspect penstocks by either building scaffolds inside the tunnel to climb through the tunnel or swing down from the gate in steep tunnels such as Glen Canyon Dam, AZ. This is a labor and time demanding practice and poses significant danger to the maintenance personnel.
In this work we replace the human inspection personnel with an autonomous Micro Aerial Vehicle (MAV) that can collect high-resolution imagery from inside the penstock autonomously. The inspection process can be completed within tens of seconds even in huge dams with a moderately trained operator. The operator's role in this scenario is just to prepare the robot for the flight and give high-level commands such as take-off, land and inspect.
This study is a demanding work and requires significant engineering effort for custom-designing an optimized platform, electrical design and software development. Furthermore, to our knowledge, the current literature does not provide solutions for state estimation in symmetric and featureless (visual and geometric) tunnels. In this work, we present a complete system design with all software and hardware components designed and integrated for this specific scenario.
You can find related videos which extensively demonstrate the inspection process and the underlying algorithms in the video lists below.
Also in the publications page, you can find our related work.