ENDURANCE

West Lake Bonney, Taylor Valley, Antarctica
Reporting from East Lake Bonney Basecamp

Mission Objective: Southwest quadrant bathymetry cleanup.

The main portion of the team was up at 8:30am and off to the bot garage. The late team (who planned yesterday’s mission at 3am) of Chris and Bill slept in till 10:30am then came to the lab at 11:30am along with Marianne Okal from UNAVCO. Marianne would be running a series of LADAR scans of Taylor Glacier and the Bonney Riegel narrows to complement the underwater 3D map. Together these would form a permanent 3D temporal record of the geo-spatial topography of these features at this point in time.

Marianne Okal sets up for another LADAR scan at Taylor Glacier. The far tripod, with its white sphere, is part of the geo-registration system. Three of these spheres, whose GPS position is known, are positioned within each scan frame and the laser ranges to these spheres form a unique registration solution to the position and orientation of the LADAR. Like the underwater 3D data, we use UTM (Easting, Northing, and Elevation, in meters) as the reference coordinate system.

At noon there was an exclamation from Shilpa at Mission Control that the internet was back up… following another 7-day absence. There was a momentary stuttering halt to the mission checklist while everyone dashed to check their email. By 1pm all of the system checks had been cleared and the bot was underway on Bathymetry Mission #5. The mission plan (see figure) was as similarly complicated as the previous mop up missions, with the effort being to fill in the gaps from the previous surveys. In this regard it is worth pointing out that there was no apriori “most-efficient” method currently developed (by us or anyone else) for bathymetry mapping of an unexplored sub-glacial lake. The approach we took in 2008 was that the Sonde chemistry casts took highest priority; the initial scanning and exploration of the underwater entry of Taylor glacier into West Lake Bonney second priority; and, finally, to fill in bathymetry and improve the extremely limited knowledge of the lake bottom topography (obtained by the few wire drops that had been performed before 2008) as a “target of opportunity”. We took the philosophy in 2008 that we would mount the multi-beam imager in down-look mode and take what we could get in the form of bathymetry while otherwise getting the Sonde data. Fortuitously, we now knew from the successfully completed 2008 sonde missions what the topography looked like and were now able to go about a concerted program to fill in the gaps. The approaches we have selected to the November 30, December 1, 2, 3, and 4 missions have sought to fill in the gaps with the least number of missions. These have resulted in the “gerrymandered” shape to some missions, today’s included.

Bathymetry Mission 5, on December 1, 2009 was another “gerrymandered” convoluted route to fill in gaps in the lake bottom topography left by prior missions. The total mission length was 3,352 meters with an under-ice transit time of 3.6 hours (total of 4.6 hours in the water).

By 2pm the vehicle was at waypoint BA47. There was no surface tracking this time although we did keep the data link up. The only human input on the mission was to adjust the multi-beam range and gain settings to gain the maximum low-noise bathymetry data. This “supervised autonomy” approach worked well (as it has in the past) and the swath widths (and fill factor in the targeted regions) were maximized throughout the mission. By 4:12pm the vehicle was back at the melt hole, locked on, and auto-surfaced. We then submerged again and returned to F6 for further testing of the DVL-assisted auto-sonde drop code that Chris had developed. By 6:16pm the bot was back on charge on its carrier sled in the lab and the data were being downloaded. Bill completed the mission plan for December 2 and this was distributed for comment. Since this would again push the system to its limits (through the Bonney Riegel narrows) efforts were undertaken to balance the battery stacks and achieve maximum accessible charge. By 8:06pm the batteries were out of the vehicle and being manually balanced; by 8:37 pm the imbalances had been corrected and the full stacks were put on a 3.25 amp charge for the night. The canisters were left open for a final balance in the morning.

Supervised autonomy (with mission control actively modifying the range gating and gain on the multi-beam scanner) resulted in a nearly 100% fill factor in the targeted gap regions of the map. Extreme under-keep shallows prevented absolute coverage in the far western section of the map. In some places less than 2 meters of clearance under the vehicle existed—the lake shallows up considerably west of the A4-C4 line.

Going for maximum performance for the December 2 Narrows mission, the individual battery stacks are manually balanced prior to full charging.

Reporting by Bill Stone

West Lake Bonney, Taylor Valley, Antarctica
Reporting from East Lake Bonney Basecamp

Mission Objective: clean up bathymetry gaps in the northwest quadrant of West Lake Bonney. Conduct instrument alignment patch and crossing tests.

The team was up at 9am (less than 6 hours sleep for Chris and Bill) and back at the mission planning task. At 10am Kristof delivered decimated data sets for the Narrows and the southwest quadrant shore scans. These were imported into the mission planner and work continued until 2pm. Bathymetry Mission 4 was a complicated, “gerrymandered” trajectory designed to accomplish the dual tasks of filling in the gaps in the 2008 bathymetry data in the northwest quadrant and, as well, to acquire data that would be used to calibrate local yaw, pitch, and roll for various critical mapping instruments. The main multi-beam imaging system was moved from its forward-looking mapping and obstacle avoidance position to a down look orientation for swath bathymetry mapping.

Kristof (left), Vickie, Rachel, Shilpa, and Chris go over the November 30th mission plan.

Prior to November 30, 2009 this was the plan extent of the bathymetry that has been acquired for West Lake Bonney and represents a composite of all 2008 and 2009 missions to date that collected bottom topography information. The blue polygon-enclosed areas represent gaps in the data set. Our goal was to fill these in to the maximum extent and leave no gap greater than 25 meters.

The gerrymandered mission of November 30 was meant to sweep out all of the blue polygon gap areas shown and, as well, to conduct calibration tests that would acquire data needed to align the critical mapping sensors in yaw, pitch, and roll relative to the vehicle coordinate system. The “patch” test used segments BA22-BA23 and BA24-BA25; the “crossing” test used segment BA35-BA36 in conjunction with BA30-BA31 and BA32-BA33.

By 2:40pm the bot was down hole and on its way. As soon as it cleared the positive buoyancy test Vickie disconnected the fiber and the next fully autonomous mission was underway. Kristof took a turn out on the ice with tracking the bot in real-time. Five hours later, having run the maze-like mission, the bot found the melt hole, locked onto the light beam and surfaced on its own. The entire mission had proceeded flawlessly with a run time of 3.8 hours underwater (4.8 hours total in-water mission time). The total underwater traverse was 3,481 meters. We had, using the mission planner, conservatively predicted 27% power reserves at this point; the actual reserve was 35% at surfacing.

Vickie breaks the data link to the vehicle for the second day in a row. From this point forward the vehicle was fully autonomous.

Given that we still had substantial power remaining, the vehicle was re-ballasted for neutral running at 5 meters and sent to station F6. A series of cast tests were conducted there (fully automated drops) using a simple control filter comprised of the Profiler encoder, the Sonde altimeter, and the Doppler vertical ranging. We will later include the following additional auxiliary sensed data to the control filter: bot depth sensors (meters); sonde depth sensor (meters); nadir obstacle avoidance sonar (meters to bottom); and central multi-beam range (meters to bottom).

At 8:54pm the bot returned to the melt hole, locked onto the guidance light beam and rose up the hole. The bathymetry data were downloaded and subsequently plotted later in the evening. These revealed some curious behavior in the data. In the central portion of the lake we obtained nearly 100% of the intended topographic fill content we were seeking. However, in the shallows along the northwest shore the acoustic reflectance characteristics of the lake bottom were significantly different, and the valid range returns from the multi-beam comprised a narrower swath than was originally expected. The instrument in question had software selectable range gating and signal gain adjustments, but these are frequently sensitive and too much gain, which would have led to a wider return swath, would have resulted in noisier, less accurate measurements unless the range gates were appropriately set. This process, currently, is something best done by a human while observing the data feedback. It would have been caught immediately had this been a human-supervised autonomous mission, as we had until now been running. It pointed to the limitations of a robotic system in a purely autonomous setting. Silicon intelligence only responds to that which it is programmed to perceive plus a few simple behaviors driven by direct sensed and derivative data.

Bathymetry fill results from the November 30 fully autonomous mission. Nearly complete fill was achieved at the center of the lake; less coverage (but still acceptable) was obtained in the shallows towards the northwest shore, largely because of a reduction in the acoustic reflectance of the material in that area.

Shilpa retrieves the iUSBL transponder prior to arrival of the bot. Since the data fiber was disconnected, we tracked the bot position using the through-ice emergency beacon. Kristof, who was on duty, radioed the message to Shilpa who pulled the beacon just as the bot arrived under the melt hole.

Returning home, image 1 of 3: the bot crosses the melt hole at a depth of 5 meters.

Returning home, image 2 of 3: the bot locks onto the central guide light beam.

Returning home, image 3 of 3: the bot surfaces, still locked onto the central beam.

Reporting by Bill Stone

West Lake Bonney, Taylor Valley, Antarctica
Reporting from East Lake Bonney Basecamp

Mission Objective: complete the final section of lake edge scanning with an autonomous, fiber-disconnected mission (we will retain the physical fiber tether but the link will be broken at the bot house). Subsequently, perform several autonomous sonde casts at F6 to test new code that has been developed since the November 21 and 23 instrument pod grounding incidents; then a calibration of the sonde encoder.

The crew was up and eating breakfast at the Jamesway by 7:30am. A full team meeting was held at 9am to plan out the final 2009 missions. Most of these focused on the completion of the lake bathymetry and a notable re-run of the Bonney Riegel narrows mission—not because of data error but in order to gain a second, temporal, data set that would show how the chemistry measurements changed at the shallow sill separating East from West Lake Bonney (WLB) as a result of the now increasing melt runoff from Taylor glacier. The team agreed to the following remaining missions:

• Nov 30: tether-disconnected (at bot house) bathymetry, northwest quadrant WLB
• Dec 1: tether-disconnected (at bot house) bathymetry, southwest 1/3 of WLB
• Dec 2: Narrows sonde re-run, with data fiber and with manually supervised sonde drops
• Dec 3: Narrows sonde re-run (if necessary), tether-disconnected (at bot house) or mop-up bathymetry southeast quadrant behind the limno cables
• Dec 4: pack up

The mission plan for November 29th, 2009 was fairly straight forward—a bathymetric side-look scanning run at 80 meters standoff range down the southeast edge of the lake to stitch together the previous two lake edge scans. But it involved moving through the limnological LTER zone. For security we left the data fiber attached to the vehicle (but not connected to mission control), so we again used the pivot point (tube) west of E10 to ward the fiber away from the limno cables.

The entire crew was on the way to the Bot Garage at 9:45am. By 12:30pm the mission was underway and ice-picking positive buoyancy checks At 12:35pm Vickie disconnected the data fiber. The bot was on its own. In order to keep some measure of what was happening Kristof and Bill proceeded to follow the vehicle using the thru-ice tracking system. They proceeded directly to the E16 sonde cast location (using GPS to re-locate it since the flags had already been retrieved after the surface fix was achieved). The vehicle proceeded smoothly down the southeast shore to B8, which was well behind the prior limno “no fly” zone. We were again using the PVC pivot tube (small red circle about 140 meters southwest of E10—see mission plan figure). All proceeded according to the program and at 3:30pm the vehicle returned to the melt hole with sub-meter navigation error.

Vickie prepares to free the bot.

The deed is done and ENDURANCE is running fully autonomously, without supervision from mission control.

Results of the November 29th lake edge bathymetry mission. Only 1/100th of the points acquired are plotted here.

Chris subsequently ran a few automated sonde casts at F6 to test new code that integrated several auxiliary sonar systems to augment the altimeter on the Sonde instrument pod. The remainder of the time until 8pm was used to calibrate the Sonde depth sensor. Kristof and Shilpa read off the precise descent distance (using a metric fiberglass tape) while Chris stepped the Sonde downward via software commands. Bill recorded the encoder ticks and the pressure sensor readings. With these entered into a spreadsheet we were able to develop a very accurate compensation system with which the Sonde altimeter readings, and those of alternative sonar soundings from other instruments, could be correlated. We planned to test this new concept following the December 3 mission.

With the bot suspended out of the water, the Sonde pod is lowered to lake bottom with a metric tape trailing behind. This external reference provided an absolute scale against which to calibrate several onboard sensors including the drum encoder for the spooler, the pressure sensors on the Sonde, and the down-look altimeter on the Sonde. These would later be fused in a new filter for enhanced autonomous bathymetry tests.

The planning for the next few day’s bathymetry missions meant post-processing all of the prior data (scores of gigabytes, including that from 2008), importing those into the mission planner, and then developing paths to assure we had as close to 100% lake bottom and ice cap topography as possible. Chris wrote code to decimate the bathymetry data while Bill did the mission planning. Both were up past 3am.

Reporting by Bill Stone

West Lake Bonney, Taylor Valley, Antarctica
Reporting from East Lake Bonney Basecamp

Mission Objective: recover, process data, go on hikes, eat lots of food.

This was our “Thanksgiving” day (the real holiday had passed several days ago while we were in the middle of key missions that we wanted to get under our belt before taking another break). Everyone slept in. The crew was up around 11am. Peter worked on his December 6 lecture back at McMurdo station. Doing similarly, Bill reduced data for the narrows and the November 26th glacier exploration to generate cross sections and movies for the lecture.

Several people went on hikes—it was a beautiful sunny day with little wind, a rarity for Taylor valley. Bill and Vickie hiked 3 kilometers past the east end of East Lake Bonney. They met Shilpa and Emma on their way back after having retrieved an ATV they had left at the east end of the lake. When they returned they found the vehicle pointed in the reverse direction from where they had left it. This was Shilpa’s practical joke. The four of them rode home together on the ATV. Meanwhile, Kristof hiked to the ventifacts, descending to the east end of the lake and then jogging back to camp.

Dinner was on at 5:30pm—way early for us—and there was lots of it: turkey, gravy, dressing (both vegetarian and meat), peas, corn, roasted veggies, kumara (New Zealand sweet potatoes), croissants, and pumpkin and cherry pies, all fresh baked by the team.

Chris supervises the carving of the Thanksgiving turkey at East Lake Bonney camp.

The team digs into a fine meal. Clockwise from left: Emma, Vickie, Chris, Rachel, Peter, Kristof, Shilpa, Bill.

Reporting by Bill Stone

West Lake Bonney, Taylor Valley, Antarctica
Reporting from East Lake Bonney Basecamp

Mission Objective: Acquire lake edge bathymetry around the south side of the lake to the narrows at Bonney Riegel and then return via the entire north shore.

We were up at 8:10am to overcast skies. It was relatively warm (for Antarctica). The thermometer on the Jamesway read 7C but no one believed it since it was in direct sunlight. The wind chill on the ice felt much colder. But the sun was having an effect on the lake ice on the south side edges and our normal ATV route had become dangerous. We crossed the main bulk of East Lake Bonney directly to the north and then rode the much smoother, harder edge “moat” ice from there to the lab.

The sun had also affected the bot garage lab, ablating the edge ice and dropping the foundations on the east side of the structure. We periodically inspected this—in a single season a third of a meter of ice ablation was not unheard of here—and today found 40 mm gaps along the middle east side edge beams. Fortunately, we had known about this effect in advance and the foundation was equipped with industrial screw jacks, which we now activated to level up the lab.

Solar ablation on the east side of the bot garage foundation left a 50 mm vertical gap between the support beams and the upper structure, possibly explaining the difficulty we had been having of late with moving the loaded gantry inside the lab.

Fixing the foundation problem amounted to a few turns on the leveling jacks spaced about the outer perimeter of the lab. Here the 50 mm gap we discovered in the morning has been closed up.

By 11am the mission plan was generated, the IMU aligned, and the launch checklist well underway. This would be the most ambitious mission yet in terms of traverse length. The multi-beam sonar imaging system was again set up in side-look mode. No special physical arrangements were needed to do this. ENDURANCE is axysymmetric, so it is directionally-insensitive. So to map the side of the lake while moving forward we simply rotate the vehicle 90-degrees to the velocity vector and the onboard coordinate system transformations are automatic. At 11:55am the bot was in the water and on its way on autonomous bathymetry mission 2. At 5pm the it returned to the melt hole in one pass from the western end of West Lake Bonney without any additional search, discovered the central alignment beam and surfaced directly. The water was clear enough to see the vehicle at -5m as it returned. There were no incidents on the entire mission. There was no intervention from the team. Mission duration was 4 hours and 50 minutes with a trajectory length of 3.6 kilometers. A total of 34 million valid (non-noise / non-multipath) bathymetry measurements had been made and logged. This left only selected gaps in coverage across the lake which we felt could be covered in no more than four additional missions.

The longest mission yet for ENDURANCE was relatively simple in geometry: proceed straight to the south side of the lake and follow the perimeter around, scanning the shoreline from an 80 meter standoff radius. This resolved a long standing debate that had gone on regarding how to effectively capture the shallow portions of the lake bathymetry. In shallow water the imaging sonar, pointed in down-look mode, was inefficient because there was not much distance below the keel of the bot for the beam to fan out. So we would have had to make a very large number of concentric passes in this fashion to ensure we had measurement data over these areas. But if we inverted the problem and scanned from the side, we could cover the entire problem area in one pass. The 80 meter standoff distance was chosen because this had proven to be a maximum range for the imaging sonar before we began seeing noise in the data. The objective was both uniform as well as quality bathymetry data.

In Mission Control, Kristof (left), Shilpa, Chris, and Peter monitor, but don’t interfere with, the independent actions of the bot. The 5 hour, 3.6 kilometer mission proceeded without incident.

The crew was back at the ELB Jamesway by 6:30pm… a short day by our recent standards. Total automation can be boring sometimes.

Results from Autonomous Bathymetry Mission 2: a total of 34 million measurements along most of the lake shoreline. Internal gaps in the data will be filled in during the next four missions.

Reporting by Bill Stone