Next week, just before we visit the Blender Conference again, we'll be traveling to PTS Space in Germany and spending a week in their Lunar Regolith Lab to scan simulated lunar soil.
[Note: Images in this post are from various Apollo missions and not our textures]
The "soil" (called regolith) on the Moon is an interesting challenge for expeditions and researchers. The low gravity and lack of atmosphere creates a unique environment which, over a long time, results in a fine abrasive substance all over the Moon's surface. This is unlike any natural soil on Earth, and is the root of many challenges in designing rovers.
The regolith is, in my understanding, essentially like very fine sand. So fine, that it apparently can behave like a liquid at times. It gets easily airborne (due to having little resistance in the form of air or gravity), it likes to stick to everything (imagine cake flour), and being so abrasive it can wreak havoc on mechanical parts or camera optics if they aren't designed to endure these conditions.
This is why facilities like PTS Space exist, so that researchers can test their designs with fake moon dust, manufactured to simulate the real thing as closely as possible, before they ever leave Earth.
Poly Haven doesn't make rovers as far as I know, so what are we doing there?
A researcher from the University of Luxembourg contacted us to explain that he is working on a machine vision project to solve rover navigation on the Moon. See, there's no GPS on the Moon, so it's another challenge for rovers to navigate accurately, especially if they are autonomous.
Image based navigation using machine vision is a logical option here, however these methods rely on a well trained model to be effective, and a well trained model requires good input data. Given that there isn't really a place on Earth that looks like the Moon, how do we get good input data?
We can't exactly send hundreds of rovers to the Moon just to capture images to help future rovers navigate better... but we can create hundreds of virtual rovers on a virtual moon to create the training data.
To do this, however, we need a good virtual Moon. We already have coarse scans of the Lunar surface from satellites which can help create large terrain features, however we do not yet have great publicly available small scale scans of the Lunar Regolith.
Thus, we were invited to come and scan the regolith simulant at PTS Space :) The goal is to capture several variants and sizes of regolith textures, as well as some footprint/rover imprints, rocks, and other structures needed to render realistic Moon environments.
This does unfortunately eat into our travel budget for next year (the facility and expertise are being provided to us, but we need to pay our own way to get there), and we're risking ruining our camera gear in a uniquely abrasive sand pit, however for such a unique opportunity in good timing with the Artemis and other Lunar programs, I believe this will be worthwhile :)
We'll write a proper blog post about the trip when we get back and share all the juicy challenges of scanning for a week in PPE while trying to protect our camera gear and not screw anything up in front of people much smarter than us :)
Without your support on Patreon trips like this wouldn't be possible, so thanks again!
-Greg