Sending the crazyflie into a confined space for inspection

[nperentos]

I would like to pilot my crazyflie down a large diameter (d= 50cm) 50m-long pipe so that I can visualise how much it is damaged/blocked. RC car is not an option because there are large height changes and drops inside this pipe.
What would be the best approach to achieve this?
I already have the flow deck, the usb radio 2.0 and I plan to get the z ranger. What about fpv? Would the AI deck work with custom code? Given the price of the AI deck, perhaps I would be better off with getting a beginner drone with optic flow sensors, e.g. the cetus pro ?

[gemenerik]

I would recommend checking out existing research in the topic to start with, but I can share some ideas.

Navigating and mapping a pipe to identify blockages is challenging. Flying in a pipe is difficult on its own due to the confined space. Beyond this, you will need to develop methods to navigate and map the pipe. These challenges apply regardless of the specific method you choose for detecting blockages.

To address these challenges, some on-board positioning system is essential. I would recommend a Flow Deck, which uses an optic flow sensor. The Flow Deck is also equipped with a downward-facing Time-of-Flight (ToF) sensor, similar to the Z-Ranger, which can be used for maintaining a flying height. Note that the Kalman Filter does not deal with terrain height variations out of the box, interpreting changes in Z-reading due to terrain height as changes in altitude. This could complicate navigation and/or accurate mapping in a pipe with varying heights. Combining it with the Multi-Ranger Deck provides ToF sensors facing forward, backward, left, right, and upward. You could do some basic simultaneous localization and mapping (SLAM) using this set-up. Then you can find blockages / damages in post processing of the map.

If understanding the nature of the blockage is a priority, the AI-Deck can be used for visual inspection. Perhaps it would be easiest to store images together with a state estimate to a uSD card with the uSD-card deck, and do off-board post processing.