Intro to Crevasse Detection

I thought I would provide some more background on my work in crevasse detection. Suk Joon and Tom Lane did a great job chronicling their efforts at Summit to equip Yeti with an instrument sled and control code. However, sample acquisition is only one of Yeti's many capabilities. His other profession, of course, is crevasse detection.

What is Crevasse Detection and Why Should We Care?
Crevasses are the ice analogues of crevices in rocks. They are cracks in the glacier. Crevasses can be constrained to the surface layers, consisting of snow and firn, or can penetrate as deep as the glacier base. I am concerned primarily with surface crevasses in the top 12 meters of glacier firn. Firn is a term for a medium density layer of snow in between harder, deeper ice and softer surface snow. Crevasses in the firn and snow layers are dangerous to vehicles traveling across the ice sheet. They are particularly dangerous because they are invisible to the naked eye: the winter accumulation and storms result in the formation of a snow bridge across the crevasse mouth. Since snow is a very plastic material at low strain rates, the bridge is eventually able to thicken enough to hold several meters of snow over it's opening, rendering it invisible on the surface. If you were to look across the expanse of a crevasse field on a glacier in the winter, you would not see a single thing but uninterrupted, white cold beauty. It almost beckons you to bound forward in delight, only to fiercely swallow you up in one gulp. The snow bridge strength is rarely known, and our only insight into its thickness is a ground penetrating radar investigation.

Here are two really good but really creepy pictures of a hidden crevasse, before and after it was driven over by a tractor.


This is a picture from inside the tractor cab, looking forward. The black flags are markings for known crevasses, so the tractor was driving in another direction trying to find a way through.



This is a picture from the back of the tractor cab. You can see where the tracks are interrupted by the big hole. After the tractor drove over it, the snow bridge crumbled under the weight of the tracks, and the crevasse opened up. If the bridge had broken just one second earlier, the entire rear end would have fallen into the crevasse. This is what is known as an unpleasantly close call.

Yeti's Next Phase: Autonomous Crevasse Detection

As 2011 closes, Yeti has made some serious improvements in both his behavior and abilities. Our exiting hero, Suk Joon is off to serve his two years in the military. We're not sure where Tom is, but we hope to see him soon, and thank him for all his work on Yeti for his thesis. Danny Dumond has also woefully departed, leaving the last remnants of terrain classification and Yeti for some other at some other time. Danny has successfully started a position at Aptima, a government defense contractor. She will be doing research and apparently lots of proposal and grant writing. Good thing Dr. Ray isn't there to correct your papers! (just joking, my humblest apologies, it's all in good fun). Good luck, our Departed Danny Dumond!

And now, I suppose, our dear Yeti and I are left to bond, hopefully without too much struggle. Yeti's next phase is now my task and thesis: A Robotic system for automatic crevasse detection using ground penetrating radar. Yeas or nays on that as the be-all end-all title of said thesis?

Yeti's newest capabilities include more complex search designs (rosettes and grids), real-time interface with the ground penetrating radar, a SIR-30 from Geophysical Survey Systems. And last but not least, the pièce de résistance, real-time classification of GPR scans for detecting buried crevasses. Of course, a huge thank you to GSSI for their enthusiastic and generous help. They supply us with Ground Penetrating Radars that Yeti tows on his backside. Here's a simplified diagram of his new get-up.



Some sadder news: currently the US is at a stand-still in terms of its government-funded research and logistics in Antarctica. This is because the US Antarctic Program (USAP) does not actually own a functioning ice-breaker. We've been using kind Sweeden's, but this year they need them all for various reasons, and the US is up an ice stream without an ice-breaker, literally, pun intended, etc. What other frozen countries up there might be able to lend us a hand? Ah yes, privyet Russiya. We have asked to borrow just a teensy one, since they have so many and all, but there's a bit of a problem. It seems that the Antarctic Treaty does not allow nuclear weapons in the South Pole, and does not particularly expire. And of course, you guessed it, Russia's ice-breakers have nukes on them. Sigh. The upshot of all that is, that those high up above are trying to come to some sort of agreement on the unfortunate mess. Here's the link to the New York Times article.

This problem is especially disconcerting to Yeti, because he is currently being funded by stealthily signing himself up for Heavy Traverses across Greenland and Antarctica funded by the USAP and the US Office of Polar Programs (OPP). But, he maintains a positive attitude, and I look forward to his training and development! Next I'll try to post a movie of the crevasse classification system I've developed! So far, I've gotten up to 92% accuracy with Support Vector Machine and Logistic Regression models trained on actual crevasse GPR data from Antarctica and Greenland.

Yeti Press

Wow looks like people are taking a liking to our dutiful friend Yeti!

Newspapers
New York Times
Discovery
Popular Mechanics
The Dartmouth
E&E Publishing

Blogs and Such
http://www.autonomousrobotsblog.com/yeti-robot-avoids-snow-traps/
http://www.popularmechanics.com/technology/engineering/extreme-machines/yeti-and-cool-robot-the-rovers-tough-enough-to-explore-greenlands-ice-sheet
http://www.gizmowatch.com/entry/yeti-robot-detects-polar-crevasses-advance-transportation/
http://www.ecofriend.com/entry/yeti-battery-powered-robot-helps-detecting-crevasses-ice-landscapes/

Here are more links to related pages!
Photos
Bio