Autonomous Lidar-Equipped Drone Inspection Of A MultiBillion Dollar Underground Power Station
- A client approached Candrone with a challenge – create a high resolution 3d model of a 500m long, half-full tailrace tunnel for an underground and continuously operating 800MW power station
- There was no known way to safely inspect the tailrace or outlet of the tunnel due to its length, restricted access at both ends, rapid water speed, and unknown conditions in the tunnel. Inspections have not been possible since it was constructed several decades ago.
- McElhanney’s Mapping Division contracted the Candrones service team to plan, execute, and deliver LiDAR data collection of the tunnel.
- Our team used the Hovermap Autonomy Level 2 capabilities, and the Matrice 300 RTK to autonomously navigate and map the tunnel beyond the visual line of sight.
The underground power station is situated in a remote part of British Columbia, and has been continuously running since its commissioning in the 1950’s. The owners have long wanted to inspect the tailrace tunnel to plan pre-emptive maintenance, however until now there has been no safe way to do this. The tunnel is 8m wide, goes 500m into the mountain, is up to ½ full of rapidly flowing water, and has a bend 150m from the outlet, making the final 350m stretch to the turbines beyond line of sight, and outside radio range. Furthermore the only viewpoint to the tunnel entrance was 200m downstream at a bridge, with a 300KV transmission line in between, and low hanging trees covering the entrance to the tunnel.
Planning missions from the office often presents a different reality when you get on site. We often bring out a "scout" drone when flying in unfamiliar environments, which is generally a smaller and less expensive drone like a DJI Mavic. This enables our team to scope out the scene for unforeseen obstacles and hazards without risking our expensive equipment.
This mission was made more challenging by the high voltage powerlines within the operational area. The lines can produce electromagnetic interference, which causes compass errors in the drone. Furthermore, GNSS errors often occur due to flying under tree canopies and between mountain valleys. We resolved these issues by staying away from the transmission lines as far as possible and disabling the drone's GNSS.
Battery management is critical to the safe return of the drone and payload. The benefit of the Hovermap and AL2 capability is that it will automatically return home even in GPS-denied environments. The connection did drop out after the bend in the tunnel but recovered once it had returned to line of sight. The DJI Occusync link provided a more reliable first-person view (FPV) camera feed and control links between the drone and controller even past the bend in the tunnel, which helped check the status of the aircraft when we lost connection to the Hovermap.
The Hovermaps Autonomy capabilities made the mission possible by providing obstacle avoidance and flight stabilization in GPS-denied environments. The mission took approximately 20 minutes to complete and a half-hour of data processing to generate a point cloud of the tunnel. Providing the hydro station stakeholders with a look into the tunnel after seven decades enables them to make data-driven decisions on maintenance for the years to come.
The Hovermap is a SLAM-based LiDAR scanner that can be used as a mobile laser scanning system or adapted to a drone for aerial scanning. Beyond its mapping capabilities, the system can provide autonomous navigation and mapping for drones.
Autonomy Level 1 (AL1): Pilot assist mode.
Pilot-controlled aircraft benefit from omnidirectional collision avoidance, creating a bubble around the drone. Furthermore, the AL1 enables position to hold and velocity control in GPS-denied environments. We used this capability to navigate around the obstructions at the front of the tunnel.
Autonomy Level 2 (AL2): Autonomous waypoints
Includes AL1 in addition to autonomous flight in GPS-denied environments using tap-to-fly waypoints on a live generated point cloud. This capability was critical to the mission as the drone was required to navigate 200 meters of tunnel autonomously due to lack of connection.