Up to Summit

After threats of a 6:30 am flight, we relaxed a little bit when we got word that the Hercules would arrive around 10:30. The whole team headed out to the transition to test the tow rig for the fuel bladders they're hauling on the traverse. Apparently the skis supporting the tow rig had some issues that would take a day or two to resolve. At that point, we heard that our plane could be delayed for a day or maybe two, who really knows? Kevin and I tried to fight our drive to escape the toast and mashed potatoes that have been sustaining us here, and headed back to while away the hours on the interweb and take a crack at final cut pro for video editing.

After a couple hours, we heard news that the C-130 was indeed on its way and would land in several minutes. We ran to grab the last of our gear from our hangar space. As we saw Yeti for the last time, Kevin and I chose the good cop/bad cop approach to behavior shaping. Yeti is easily anthropomorphized since it's behavior can't always be explained. Jim hopes to run it for a few good short data runs, then run it until it into the ground.

C-130 Landing


We waited on the Tarmac while the crew removed the Jet Assisted Take Off units from the back of the plane. We then loaded into webbing benches strapped to the wall and stuffed in earplugs. The Herc flew south for a while before heading east inland towards summit. The sea ice stretched as far as we could see. We're intrigued by the possibility of hydroplaning a snowmobile from ice berg to berg, but that'll have to wait until our next trip up here.







Unloading at Summit


Our experience at the Summit consisted of running to find the 'pee flag,' then standing in prop-wash for 45 min as the plane unloaded fuel for the camp. We were happy to get back to Kanger, though apparently nobody figured out that we were supposed to eat. We got a couple of dehydrated meals from the crew here and headed back to our dorms in the science support building to boil some water and enjoy some subsidized scotch. Kevin and I are realizing our four days here could be pretty busy with writing, but we're planning to take a day and sleep in tomorrow.

Fuel bladders on sleds ready to go

Yeti on its own

After our successful run yesterday, Jim was convinced that the robot was prepared for him to operated it on the traverse. This is great news because it means we'll get good radar data from crevasse zones, which were further out than we had the opportunity to go on this trip. We're also excited because it means that Kevin and I won't have to worry about getting Yeti home, which was a massive undertaking on the front end of the trip.

Today was our last day in Thule, which we spent the day teaching Jim the intricacies of Yeti's internals, including the embarrassing workarounds we were forced to make when working on electronics in the field. The system is robust though, and we expect that it will perform well once its our of our hands. We won't know for another 6 weeks or so, until the team gets back.

We celebrated by going to the gym and getting our our aggression with an automated tennis ball launcher inside a squash court, after getting tired on the rotating tread climbing wall, which we followed that with some Scotchernetting.

We'll be off the grid for a day or two, after which we'll hopefully have some videos of Yeti. We're both excited to get home at this point, see you all soon.

On the Cap

We spent most of yesterday resting up and preparing to head out further up the glacier than we've been before. After a bunch of test runs yesterday, Jim was suitably convinced that yeti was up and working well enough to take it out further to run a course through some of the rougher sastrugi on the traverse route. We drove out to the third waypoint, towing Yeti on a sled behind the snowmobile and running the Tucker Sno-cat along side.

When we got to our test location, we got Yeti ready to go and pressed play. Pretty soon, it tacked north and veered about thirty degrees off of it's intended course. I groaned as Jim seemed thoroughly unimpressed before we manually drove it back on track and pressed play again. Within thirty seconds, it turned north again off of our track, and I started to bite my nails. We knew our controller worked from all of the other successful tests, so I guessed that our controllers might have drifted from being bumped around for on 10 miles on the drive out. We did a quick field re-calibration and held our breath. It worked perfectly, and Yeti drove off towards the first waypoint.



This exceedingly boring video gives a quick sense of speed and scale and records the first autonomous GPR record in history. That's a lot of qualifiers but we're excited.



The snow was soft and the sastrugi was relatively large, but Yeti handled it all. It did get stuck several times when it hit large but soft features at just the wrong angle, but that problem was the result of our small tires, and won't be an issue once we replace them.

The radar that we got back looked pretty good to us, though we're waiting to see what the experts say. It looks like the radio on Yeti has some interference with the radar, but that won't impede its detection of crevasses and can be turned off once we're done with testing and out of the prototype portion of the project.

Kevin and I are actually excited to be heading out at this point. The rest of the team has been great but we've been holed up for some absurdly long nights in the nerdery, and ready to see the sun go down again. We'll head up to the icecap summit on Tuesday on a hercules C-130 cargo plane, and figure out the plan from there. I haven't had time to cut all of our video yet, but I'll work on that and hopefully have it up by the beginning of next week.

Right before heading up the ice, yeti in tow


Kevin on his own


This was all Jim's idea, from a paper concept in 2004. He's glad to see it running at last.




Our workspace may not seem so bad from the pictures, but we're working 100 yards from a runway in a building in which they operate heavy machinery with diesel exhaust, which also houses a firing range, and next to our little aluminum pallet that we work on is a shelf of stacked HAZMAT barrels. We're lucky we're not out in the cold though, and after this, nothing can distract us.



Kevin seems to have a ridiculous fondness for his new ski goggles (seriously)

Video

Here's a montage of driving out on the ice.

Preparations

The last couple days have been some of the most frustrating but productive that we've ever experienced. Kevin and I just finished a 23 hour day, and have gone almost completely asynchronous from a normal schedule, and we've shifted to working from 11 am to 3 am. The beauty of doing this here is that we never have to experience the sinking feeling of working through a sunrise. Since most of our meals here are entirely identical, we have almost no reference for what the current time would ordinarily imply.

We've been furiously working to bring telemetry on line for our next test and have finally succeeded in getting the radio to talk to our computer. Yeti now sends back it's coordinates, time, GPS quality data, and it's next waypoint. What made this so frustrating was that when we added telemetry functionality to our working controller, the robot would wander in loopy circles and not seem to care about its next waypoint. This stopped being funny after hours of diagnosis revealed no major clues as to its cause, and we were only able to get it working properly by reducing the amount of data we sent. The system performs very well now, but without much of the code we've written.

We've also hammered out our last issues with our ground-penetrating radar, and have gotten that working perfectly. We expect that our next run will yield high-quality GPS-tagged radar data. Thanks again to GSSI for letting us use one of their radar units. Without that we'd never be here in the first place.

Today, we're preparing to run Yeti tomorrow on it's last major run, further out on the ice sheet. This will be a more complicated course over sustained and gnarly sastrugi, and we plan to be entirely hands off from start to finish. Yeti's batteries are our biggest concern now, as several of them have broken on us and only 5 1/2 out of the original 9 are working. Yeti should be able to finish the several mile course, but it'll be close.

We heard yesterday that we would have to leave three days earlier that expected, though we'll be getting back two days later! The only way to get home is by flying to the Greenland summit station and spending two days there, then flying back through Kanger which requires a two-day layover as well. We won't have access to Yeti so it'll be a mandatory rest that Kevin and I could probably use, though we'll miss GreenKey weekend at Dartmouth by a day.

More pictures on the way, stay tuned for tomorrow's results.

Field Day

On Tuesday we took Yeti back into the field for a surprise run. We expected to have the day to work on our code and perfect its autonomous behavior but the rest of the team was heading into the field and with great weather we had to try out our first fully autonomous long-distance run.

From the base of the transition, we pressed 'play' and watched as Yeti climbed up the glacier towards its first waypoint about 3km up the icesheet. Kevin and I struggled to keep up, bogged down with the toolstation and backpacks full of gear. Yeti climbed through the sastrugi field with the largest features about 2.5' tall without getting stuck once, holding its bearing well as it bumped around the rough terrain.

About 2 km up the glacier we heard Allan heading up the glacier on a snowmobile, and he told us we'd have to cut it short since people were heading in for the day. We hoisted Yeti up on the crane from the Case tractor and brought it back to base. We spend the rest of the night coding, logging a 17.5 hour work day.

Our radar record showed us very little about the ice surface since we apparently had the gains set such that it ignored most of the first layer. Allan showed us how to set the radar accordingly, which we changed for Today.

Today we got up again at 7 and packed up to head out to the field again. This time we decided to use a snow machine, and found out how ridiculously helpful it is. Running with a laptop on ice has proven pretty difficult, but laptoping on the back of a snowmobile worked out extremely well.

This time, we pressed play and Yeti repeated the climb, this time making it the entire 3km to the first waypoint before turning around and heading to the second point at the bottom of the glacier. The radar data looks great as far as we can tell, though we'll have to take a closer look at it when we have time tomorrow.

We've learned a lot about the system at this point and are pretty close to the point where it would be viable to use in conjunction with the sno-cat for the GPR work on a traverse. We've identified the things we need to improve to get it to the point where it could do the majority of the survey work. We've got our hands full updating the code, but we're encouraged by the results and have had great success so far. Our next major milestone is a fully hands-off long range run with GPS-tagged radar data, and we plan to do this run on Friday.

Yeti with mount Dundas in the distance




Looking out over the fjord filled in with sea ice in the distance

A mind of it's own

Today is exciting since we finally achieved all of our goals of the original project.

After a brutal day of coding yesterday, in which Kevin and I spend 10 hours figuring out how to send a number over the radio from the yeti to our computer, we conquered the ridiculous. What kevin thought would take him half an hour ended up taking the whole day. The details are not as interesting as the fact that this allowed us to troubleshoot the control system much more easily.

This morning, we fired up Yeti and fed it a GPS coordinate out on the Thule runway. Technically I think we're breaking a lot of rules by driving around and taking photos in a controlled area, but we haven't been shot yet so it's worked out. The first time we tried this, Yeti turned and drove away from the coordinate. A couple more runs revealed that it was not accurate but consistent. Kevin figured out that he had a sign error in his code which would make it run in the opposite direction. After reprogramming it, we gave it another shot and Yeti locked onto its target immediately and drove straight to it.

We spent the rest of the day making more improvements to the motor controllers which will extend yeti's driving rage significantly and eliminates the purr the robot used to make when stopped.

We'll be out in the field tomorrow so we'll have some more interesting pictures and video.

The Icecap

I'll put the pictures up front, if you care about the details of Yeti's performance on the Icecap make sure to read to the end.

After pulling an 18-hour workday on Friday (sorry for not posting), we got Yeti ready for some testing on the ice. We were going to keep working on autonomous control but it became clear that that wasn't an option as everyone we came with really wanted to see it run around on the ice.

We drove off the base for the first time, up the 12 miles towards the ice cap. We passed the Thule Ballistic Missile Early Warning Radar System for which the Thule base exists, that gives us early warning in the event of Russian missile launches. The rest of the team have been setting up for the traverse at the base of the transition onto the icecap. A large section of the transition was groomed to give ample space to tow the fuel sleds, but this gave us a path to get down to the base of the ice sheet.

Me about 200' from the edge of the ice, which isn't visible.


Kevin in front of the Tucker. This beastly machine won't be necessary the next time NSF runs a traverse.


The Tucker with the frame assembly on the front for GPR surveying.


Part of the traverse cargo in front of the transition. This ramp of ice leads up to the icecap, the second-largest body of ice in the world after Antarctica.


Thule BMEWS



Kevin had a hellacious time trying to calculate a bearing from GPS coordinates with the small computer onboard Yeti and we're going to need another day of development before we'll be able to operate fully autonomously. I reworked the driving interface which gives us much better control and makes driving a lot of fun.

Kevin commented that it's probably the nerdiest hick sport imaginable to go out 'robot rock-crawling' but we justified it by calling it testing.

Yeti performs ridiculously well on rough and steep terrain and has exceeded our expectations in its ability to get over obstacles. When we took it out to the ice cap today, we were able to navigate sastrugi that was several feet high, though Yeti high-centered and got stuck on some sculpted blue-ice features that don't occur further up the glacier.

We drove on 2/3rds of our batteries, and this lasted for about 2.5 hours of driving, 1.5 km of which was up a 10-degree slope.

Overall, we're quite happy with how it performs and we'll be doing additional characterization of the power and drive system in the next week.

Running on Blocks

We spent most of the day today working on code. This is the more interesting stuff for us, but harder to write about and less exciting for everyone else reading. We had an extremely hard time dealing with the slowness of our serial connection with the computer. The consequence of this was that every once in a while, the connection would drop the 'stop' command and the robot would keep moving. This was concerning for obvious reasons, so we spent an inordinate amount of time reworking the code to quash the issue.

We also got our GPS up and running and saw 4 satellites for the first time, and Yeti was able to interpret the GPS info for the first time which was exciting. We're now where we would have liked to be about 2 weeks ago, but we've been working around the clock as it's been.

Tomorrow we'll take Yeti outside to test GPS navigation and the new manual driving code. We only have one day to get up and running before we head up to the ice sheet transition on Saturday.