First test flight in the Central Storage Canal: insights from a challenging drone mission
Initial situation and drone system
A DJI Mavic 2 Pro, provided by FH JOANNEUM and AIRlabs Austria GmbH, was used for the first test flight in the Central Storage Canal (ZSK). Upon entering the outlet, it was apparent that only small amounts of rainwater were being carried in the canal. Deposits were minimal, as regular operation is not scheduled to begin until mid-2021 and only rainwater will enter the canal until then.
Starting position and surroundings
About 150 metres into the ZSK is the fork between the inlet to the sewage treatment plant and the outlet. The drone was positioned on this dry side arm, which is normally used for overflows. Above this area is a building with small windows and overflows, which means that the lighting there is much better than in the tunnel section of the canal. Further into the ZSK, there was about 7 cm of water, which made flying even more difficult.
First test flight: stable take-off, unexpected end
The first flight led from the starting point to the confluence with the Mur River. The video image was recorded in its entirety, and no difficulties arose on the way out. However, upon leaving the tunnel, an automatic landing occurred because the drone was in the no-fly zone of Graz Airport. This made it impossible to take off again.
Second test flight: navigation problems and crash
After returning to the starting point, a second flight deeper into the canal followed. This proved to be much more difficult. The reason for this was the complete lack of GNSS coverage in the storage canal. Without a satellite connection, the DJI Mavic 2 Pro was unable to maintain a stable position. The GPS signals coming in sporadically through openings also led to inconsistent navigation data, causing the drone to respond with automatic position corrections – a behaviour that is particularly problematic in narrow tunnel structures.
Despite these challenges, it was possible to advance a few metres further into the canal. Eventually, the drone crashed and had to be recovered from the partially stagnant water. Fortunately, there was no discernible damage, but the test flight day was ended to avoid possible consequential damage.
Lessons learned and open questions for future drone inspections
The initial tests have provided valuable insights that form the basis for further steps.
1. Submission to the TAKE-OFF programme
It is being examined whether the autonomous application can be submitted within the framework of the current FFG call for proposals TAKE OFF 2020.
2. Coordination with Austro Control
As the drone landed automatically in the no-fly zone, it must be clarified whether it is possible to enable take-offs within the zone. Without this adjustment, future operations will be severely restricted.
3. Navigation and positioning without GPS
The biggest technical challenge of the entire mission was that the drone had to operate completely without GNSS in the storage canal. However, the Mavic 2 Pro is not designed for precise positioning in GNSS-denied environments. The intermittent GPS signals led to inconsistent navigation data, causing the drone to perform automatic position corrections – particularly problematic in a dark, structurally homogeneous tunnel environment. Although there are flight modes that allow operation without GPS, these did not work reliably in the test. In addition, the Mavic 2 Pro’s purely optical distance sensors cannot guarantee stable navigation in poorly lit areas. The combination of these factors ultimately led to the crash in the second test flight. Another disadvantage for research purposes is that this model does not allow access to detailed log data.
⇒ An alternative drone platform is therefore necessary.
4. Requirements for future drone systems
The following requirements must be met for future autonomous indoor inspections:
- Indoor navigation without GPS
Pre-programmable waypoints
Live transmission of image and flight data
Reliable data communication and control
One potential option is the Elios 2 from Flyability, which is specially designed for tunnel and inspection flights. Open question: How well can it be used autonomously?
5. Amplification of the control signal
Once the drone and operating concept have been finalised, signal amplification along the entire length of the tunnel – for example via Citycom – must be evaluated.
6. Definition of the data to be transferred
Finally, the following must be determined:
- What data is necessary? (Image, sensor values, telemetry, etc.)
- How will the data be transmitted?
This decision has a significant impact on the choice of system.
Conclusion: An important step towards autonomous tunnel inspection
The first test flight showed that drones offer enormous potential in underground infrastructures, but require specialised technical equipment. The tests made it particularly clear that classic GPS-supported drones are unsuitable for inspections in GNSS-denied environments. Without reliable positioning and robust indoor navigation, neither the necessary flight stability nor the required safety can be guaranteed.
The findings now provide clear pointers for the further development of an autonomous drone solution for tunnel and sewer systems. With the right technology, safe, efficient and fully automated inspection of such facilities could become possible in the future.