Field Notes from Mauna Kea

Oct 28, 2008

Flying. All day. We started the day in Pittsburgh, PA before dawn and ended in Hilo, HI after sunset.

For us it was a long day, but for Scarab it had been a trying journey. Scarab left Pittsburgh a week earlier under the care of the 911th Air Force Reserve unit. The crew made several stops to pick up additional equipment from other groups (significantly more cargo than initially indicated). During this time, the C130 had problems causing it to turn around twice. Despite these difficulties, the crew pressed on and Scarab landed only a few hours before us.

Oct 29, 2008

Today we did all the prep work down in Hilo. We got the rental vehicles, did some shopping, unloaded the pallets from the cargo plane and performed basic checkout of Scarab. Late in the afternoon we had everything together and drove it all up the mountain.

The drive out to Mauna Kea took about an hour, with scenic views the whole time. We unloaded everything at Hale Pohaku (HP) - our home for the next two weeks. A PISCES organizer met us there and showed us how to use the forklift. With the last rays of sunlight we decided to make a final push to the field site. The road in was steep and slippery but we got Scarab in safely and left it in one of the tents waiting for us.

The site was clearly not ready; most of the tents were not prepared (including our large one), the road was not suited to many vehicles driving across it daily, vegetation covered the test site and the satellite dishes were not online. Within a few days these would eventually be fixed, but we had precious little testing time as it was.

There are only two access roads into the site; the "high" road which is steep and sandy and the "low" road which is rough. We entered on the "high" road and tried to exit that way too, but in the darkness we got stuck in loose sand on a curvy section. Both of our cars got stuck here so we hastened to the other road. We were met by John (another PISCES organizer) who had just come this way after getting stuck on the low road. With night setting in and the unappealing prospect of walking home we made one final push. This was a bumpy ride, but we made it out eager to get to work tomorrow.

Oct 30, 2008

Our first day of testing was a rough one. In the morning we moved the remaining crates from HP to the field site. After that we ran into technical problems almost immediately.

First our SPAN unit (a commercial position estimation system from Novatel) started acting up. We have been using an Omnistar subscription to improve GPS accuracy, but the SPAN now refused to use this data. While adjusting settings to the unit, the hard drive crashed. Initially it appeared that the hard drive failure was related to the SPAN, though we later learned this was not the case.

By noon these issues were sorted out to some extent and we moved on to other details; mounting the TriDAR scanner, checking all the sensors, trimming and weighing the vehicle and of course eating lunch.

In the late afternoon we started our first mobility test. One of our goals for the field season is to compare the Michelin lunar wheels with commercial skid loader tires. An important test is the drawbar pull. This involves driving the vehicle straight while pulling back with a known force. In the field, we can do this by making Scarab drag a sled and varying the weight of the sled by adding sand bags. A load cell is used to measure how much force the sled exerts on the vehicle.

As this was our first test, it took several hours. Just finding the right spot took a while; we needed over ten meters of level ground with uniform soil and no rocks or vegetation. We also had to get accustomed to the procedure; synchronizing logs, acquiring video, preparing the total station, the naming convention, double checking data streams, even the details of adding sandbags were all discussed. Add to this a rain delay and curious bystanders, and you have one long experiment.

At the end of the day we finished our first experiment. They didn't go particularly smoothly and some of the data was suspect, but we were confident in the procedure. Over the next two weeks we have many more drawbar pull tests planned and we now have a reliable and repeatable process.

Oct 31, 2008

The first order of business was to address some of the problems from yesterday. Some people helped the PISCES crew with camp setup while others worked on Scarab. We spent a lot of time on the phone with Novatel and Omnistar and identified a few problems. With corrected settings and our Omnistar subscription updated, the SPAN started working reliably with good (12cm) precision.

The afternoon and evening was spent with initial navigation tests. Scarab autonomously drove through camp avoiding crates and tents. We also sent Scarab through sparse vegetation and it behaved well. Also we verified our ability to drive in darkness with an evening run. In total we drove over 2 hours and 20 minutes covering over 260m.

In addition to autonomous navigation we also did checkout of the velocity camera, precise motion and TriDAR. The TriDAR sensor uses a laser to produce 3D models of the terrain in two different ways; time of flight and triangulation. The time of flight mode is used by Scarab to avoid obstacles while the triangulation mode is used for drill site selection (it takes longer and is not eye safe, but gives improved performance). For precise motion we drove measured distances and fine tuned the wheel radius to compensate for dead-reckoning errors. After this initial calibration we achieved <1% error on distance. Finally the velocity camera (which estimates velocity by watching the ground go past) performed poorly in measured trials. The telemetry gathered on these runs will help calibrate the sensor.

Our last activity was a simulated drill site approach. We marked the vehicle position with a physical object representing the drill hole. In darkness we commanded the vehicle to go to a distant waypoint and return. Upon completion we manually acquired scans, identified the marker and commanded precise motions to reach that marker.

At the end of the day we settled into our large tent. Today we successfully demonstrated autonomous navigation in different regions of the field site. We also completed our dry run of drill site approach. Tomorrow we will return to mobility tests, trying to tick off as many milestones as possible before the payload arrives.

Nov 1, 2008

Today started out rough. We were greeted at the field site with a blown over tent. Then right after turning on Scarab the hard drive failed again.

We replaced the hard drive in an intricate hour and a half operation. Scarab's internals are very tightly packed making access difficult. It took a full hour just to access and swap the drive. Then it took another half hour to load all the latest autonomy software.

We didn't know if this really was the cause of our computing problems, or how long until they reappeared. After a phone call back to Carnegie Mellon a spare computer was on its way. Another possible problem was thermal, the temperature was only 6 degrees celsius and likely colder overnight. Also we had already accumulated a good coating of volcanic dust all over the electronics leading to worries about electrical problems. The soil was very fine and ferrous, we resolved to try to keep this out of the rover body.

In the afternoon we did our first slope mobility tests. These tests help characterize Scarab's hill climbing ability and highlight differences between the commercial skid loader wheels and the Michelin lunar wheels. Since this slope contained rocks that might damage the lunar wheels we decided to only use the skid loader wheels today. Later when people from Michelin come, they will give us the green light to run the other half of the experiment.

Today's slope test involved driving in three modes: straight ascent, diagonal ascent and diagonal ascent with auto-leveling. After our experience with the drawbar pull test, this one went very smoothly. The exact rover motions were all preprogrammed and our procedure was now easy to follow. The test itself took less than half an hour.

Adjacent to this site is a large drainage with steep banks and loose material. We attempted to replicate the slope test here, but it was too steep for Scarab to handle. At this point we turned our attention back to autonomous navigation - could Scarab navigate through the drainage back to camp?

At 3:30 we sent Scarab down the drainage. As this area was still densely covered with invasive vegetation we had to clear out bushes ourselves. Earlier in the day we found that Scarab could get stuck in dense brush as it can not tell the difference between a bush and a boulder. An hour later we were rewarded by Scarab emerging at the base after winding its way through the drainage.

At 4:30 our next surprise came when Scarab reached some lava rocks. The robot wiggled left and right indecisively, then charged right over - narrowly avoiding scraping the bottom of the body. After reviewing the data, we found this was a calculated move on Scarab's part. It found that by lining up with the rocks and straddling them it would clear them by a few centimeters. Though it gave us pause, this maneuver was indeed faster than driving a circle around the obstacle.

We had a short break before continuing our long traverse. As night set in, we identified a number of problems related to scan merging. We didn't intervene, instead we let Scarab stumble around in the bushes several times due to these glitches. Scarab kept driving well into the night and covered an additional nearly 800m until the navigator crashed.

At 2am we left the site a little disappointed with the software bug (which was easily fixed) but the day had certainly been a success. The first slope test was a valuable data point and we clocked over 11 hours and well over a kilometer of autonomous navigation.

Nov 2, 2008

Half the crew (the ones that stayed out late with Scarab) slept in. The other half got up early for more drawbar pull tests. These tests involved 3 sites and both sets of wheels. There were some hiccups resulting in a few false starts, but overall a lot of useful data.

After these mobility tests we took another look at drill site approach. The Neptec crew was still having trouble using the TriDAR in triangulation mode, suspecting a bad fiber optic cable. They found that the cable had been connected backwards but triangulation still did not work. For now, time of flight would have to suffice.

The drill and science package arrived on site today so the evening was spent installing our payload. Mounting the drill was time-consuming but fairly straight forward. Next we turned to the science instruments, they rest inside the rear cavity on a removable frame. This should have been a simple matter, but something happened to the frame. After a few hours and some filing we got the package into the rover.

With the new instruments in place, tomorrow we hope to drill our first core sample.

Nov 3, 2008

With the new instruments mounted it was time to weigh Scarab again - 399 kilograms, nearly 900 pounds! The Norcat crew still had to do their alignment procedure. At 10 the drill was ready and we finally moved out to the drill hole.

Now it was time to practice drill site approach for real. We settled on a three phase system - first the rover would drive autonomously to about 10 meters from the drill hole. After this Scarab would 'stand down' and give a remote team in Canada direct control over a satellite link.

The remote team has three high level commands available; they can command scans, issue a precise motion of distance and bearing, and set the ground clearance. First they must locate the drill site by acquiring and analyzing a scan. Next they command a distance and bearing to bring the rover about 3 meters from the hole. Then they repeat the scan and drive cycle to position the drill directly over the hole. Finally they issue their last command to lower Scarab down to the ground.

Over the next few hours we made several practice approaches. This let us fine tune the precise motion as well as give Neptec experience identifying the drill hole and the mechanics of remotely commanding the vehicle.

By late afternoon everyone was ready for the first integrated test. First we issued a goal to Scarab and it drove to within 10m of the drill hole. Then the remote team in Canada acquired a scan and brought the vehicle about 3m from the hole. Then another scan and drive brought us right over the drill hole. After that we shut Scarab down and let the drill team take over. Drilling a core sample took several attempts and they worked into the evening.

Nov 4, 2008

We arrived early at the site and carefully removed the TriDAR. The Neptec guys could take their time inspecting the triangulation laser and get it working for the next drill site approach. They were able to trace the problem to a blown fuse and install a replacement.

While the science team was working we went down to Hilo. There are no logos on Scarab so we decided to add NASA and Carnegie Mellon logos for the media. Rather than affixing them to the shell and ruining the paint job, we decided to make a bracket for the mast.

At the end of the day we reconvened at the field site. The bracket prominently displayed the important logos and the TriDAR triangulation mode had been repaired. Unfortunately the drill had failed and the science team had not processed a core sample. This was a tough blow - they needed to push the entire schedule back a full day.

Nov 5, 2008

Since the schedule was pushed back a day, we had to squeeze in more testing time. This morning we arrived before dawn and took Scarab up to the black sandy slope north of the site. This material is very fine and quite steep preventing Scarab from driving straight up it. We got about an hour of testing done on this brand new material.

We raced down to the drill site and were ready for the final approach at 8am. After fixing a satellite problem, the remote team in Canada had their second shot at drill site approach. Then we shut down and left Scarab to drill and process all day long - again.

Nov 6, 2008

This morning we got a late start, waiting until two o'clock for the science team to finish processing their first core sample. The plan for the day was to return to the black sandy slope and perform comparative slope tests. We now have two representatives from Michelin with us, they will help guide our tests and detail the data they need to characterize the wheel performance.

Today we gained valuable insight on how Scarab operates in this material. We observed the tendency to turn uphill, the failure of the felt tread and the strength of the leveling switchback maneuver.

Even on this difficult terrain, Scarab could make steady uphill progress. We drove Scarab cross-slope and made small uphill gains. Upon reaching an obstacle we would start driving backwards, correcting the angle of attack and leaning posture to continue making slight progress. Each switchback took about 10 minutes and took us only a few meters up the slope. The whole time Scarab was fully under control and making steady progress uphill, a big victory for this chassis design.

The testing today was long and took us well past sunset but showed off an impressive vehicle capability. Tomorrow morning we're expecting to demonstrate this for some visitors.

Nov 7, 2008

We started the day with some more testing on the black sands. Our visitors arrived mid morning and we talked about rover mobility on this slope's material. The importance of the chassis, the control strategy and the wheels were all discussed.

Next we moved back to the hill west of camp. This material is a little rougher with some rocks. Now that people from Michelin were present to observe, we were eager to repeat our mobility tests in this potential hazardous area. The tread did take some damage on the sharp lava rocks, but the wheels themselves worked fine and the structure held up great.

Next we drove Scarab into the adjacent drainage and let it navigate down to camp again. Repeating this demonstration, this time with lunar wheels, was a rewarding experience.

Originally we had hoped to make an ambitious ascent up to the crater rim. This would be another nice demonstration, as well as a good photo opportunity. Unfortunately it conflicted with the plan to drill early tomorrow morning. So instead we spent the evening working with the velocity camera again, doing more calibration tests and a little autonomous navigation.

Nov 8, 2008

Today was another short day for us. We started to navigate toward the drill hole around 9am when our NASA visitors arrived. They got a chance to see Scarab do a drill site approach and inspect the science payload. Around 11am the drilling started and we left the field site.

Nov 9, 2008

Today we had our usual late start as the science team finished processing their second core sample. We finally got started around 2pm and immediately got a small scare - the computer wasn't booting. After a lot of fretting and dragging out a monitor, we learned the computer was merely performing a routine (though time consuming) self-check.

This afternoon we drove Scarab over a lot of rocks and really stressed the lunar wheels. The Michelin crew were delighted with the performance, these wheels really stood up to the jagged lava rocks.

In the evening we started our second night traverse near the top of the drainage. Two and a half hours and 300m into the mission the navigator got some bad data. After observing multiple glitches similar we added some new code in hopes of resolving the issue. Unfortunately the new code had a bug which was easily fixed.

Around 9:30pm we started up again. This time the rover drove autonomously without a hitch for over 8 hours, traveling just shy of 1km. As we approached the symbolic one kilometer mark, the right side wheels stopped moving. With sunlight creeping into the valley Scarab tried to limp along, but could only turn in place. At 5:30am we aborted the traverse and were quickly able to reset the right side.

With just a few hours until the next scheduled drilling operation we brought Scarab over to the drill hole. Around this time the wind picked up and we still had to change the wheels. We were exhausted, cold and getting sandblasted... swapping wheels took an eternity!

Today was another successful day blemished at the very end. We really pushed the lunar wheels and they held up well. On the autonomy side, we drove for over 10.5 hours and covered about 1.2km. Tomorrow (meaning, in a few hours) Scarab needs to do another drill approach and process another core sample. In the meantime we'll take the lunar wheels down to Hilo and make some modifications to improve the treads.

Nov 10, 2008

Today our team split up again. Most went down the mountain to Hilo with the lunar wheels. The rest stayed to be on hand for Scarab's drill site approach.

At the field site the wind had kicked up even more and blew over two tents. The drill site approach went quite smoothly, in the end we were only needed to turn the vehicle on and off.

Down in Hilo at a machine shop, we worked on the lunar wheels. Using some stock materials and a banding machine we added makeshift grousers. We had noted poor traction with the felt tread but adding these grousers should help in that area. Tomorrow we'll test them out.

Nov 11, 2008

Most of the day was spent testing the modified lunar wheels. We started with slope mobility tests on the black sands and noted a big improvement in traction. Scarab still had to do switchbacks up the slope, but could take a sharper angle of attack resulting in faster uphill progress.

After finishing the slope tests we went back down to flat ground for drawbar pull tests. We feared that this test might be too intense for the grousers and rip them off entirely. They did suffer some damage but held up surprisingly well. We observed significant slip between the wheel and grousers so it's unclear if they helped. We need to look at this data closely to determine the effect of the grousers.

We took off the grousers and regrouped to discuss our field test objectives. Time was running short leaving us hardly any room for more experiments. At this point we've hit our major objectives: autonomous overnight traverses, drawbar pull and slope mobility tests Scarab has autonomous navigated about 3 kilometers, quite a distance given that it has a top speed of 5 cm/s. Our mobility tests have included both wheels in multiple soil types, we also got additional data with added grousers.

Next up on the list was integrated operations with the velocity camera. In the evening we drove a short mission out from the tent and back again, avoiding some obstacles on the way. Scarab did a fine job of returning to the tent, but as with the grousers we need to do some data analysis to determine the true effect of adding this sensor.

Here is the schedule for the remaining days; 12th - drilling and processing, 13th - media day, 14th - transport to Hilo, 15th - public outreach at the science center and packing at the airbase. It's a full schedule with little room for fixing inevitable problems as they arise...

Nov 12, 2008

For the last drilling operation, the science team decided to be more ambitious. Rather than using prepared samples they opted for native soil which has significantly more moisture and may clog up the system. Also they wanted to try drilling on a slope rather than level ground for a more challenging scenario. The drill site approach was also more realistic - rather than hunting for a specific target we merely had to avoid large obstacles.

Unfortunately the operation did not go well at all. Firstly there were communication problems because this site is quite far from the satellite dishes. Once that got sorted out, the drill aborted as soon as it got into the ground. After six hours and with imminent rain we started to bring Scarab back.

While loading Scarab onto a trailer we hit a problem with one of the linear actuators that allows Scarab to raise up and down. Scarab was now stuck in the 'up' position, not particularly stable for the bumpy ride down to camp. Winding our way down slowly in the rain, we managed to get Scarab into the tent and raised on boxes. We worked into the night inspecting the actuator and got it working again.

After that close call we realized we needed to be careful with Scarab on these last few days. We simply don't have enough time to fix any of the major components and two big public events coming up.

Nov 13, 2008

A large contingent of media visited us this morning. Our group, including the drill (Canadian Space Agency) and payload (NASA) teams, discussed what we've all been working on these past two weeks. The media also covered the other groups that have been working along side us on Mauna Kea. Those groups focused on a larger scale science package suited for a lander, rather than our smaller mobile version.

Once the media left we all packed up and cleared out of camp. This happened faster than we expected and by the end of the day all of our equipment was safely out at HP. Tomorrow we will bring it down to Hilo.

Nov 14, 2008

Today went by fast. We moved everything back down to Hilo and separated what we will need at the science center from the rest of our gear. Tomorrow we'll need Scarab and a minimum of support equipment - just some charging cables, handout materials and basic packing supplies.

Nov 15, 2008

Today we demonstrated Scarab at the science center. Public response was very enthusiastic; we had lots of families come by eager to learn about our experiments. We let a lot of kids drive Scarab around with the joystick and talked about prospecting robots.

In the afternoon we packed up and moved to the airbase. We had to pack Scarab in its crate and put all of our (and everyone else's) equipment on pallets. We worked into the night until all the crates were stacked, covered and strapped to their pallets. Finally Scarab is ready to go home.