Smart tools and systems for exploring the frontier
Dom watches as the robot is lowered into La Pilita with two scooper attachments that will be used to sample bottom sediment.
The robot buried the scoopers in the bottom sediment then spun to take a sample. The procedure worked so well that the amount of sample taken made the robot slightly negative.
Marcus Gary gives the "thumbs up" after locating the bot Monday morning. After failing to resurface from a 4-hour mission the night before, the team was extremely relieved to see DEPTHX bob to the surface after Gary freed it from the rock it was stuck on.
morning dive:
afternoon charge session:
evening dive:
Marcus was up before dawn. He geared up at La Pilita as the sun came up. Honestly, finding a robot (albeit a bright orange robot) that could be anywhere in La Pilita seemed like a futile task. But, being a game soul, he descended to 75 feet and began a slow dive following the walls of the sinkhole. The water was crystal clear. Beautiful biomat-covered lobes stood out under his HID dive light, which cut through the water like a laser beam. As he was swimming, he contemplated what divers would be recruited for a deep recovery mission. First choice, Jim Bowden, is currently in Italy diving at the deepest underwater cave in the world—the Merro Well. Approximately 15 minutes into the dive, one particular lobe caught his attention. He ascended a few feet and his light illuminated the word “DEPTHX”. Miraculously, he had swum right up to the robot.
All battery power had been spent, so the robot’s lights were dark. All systems had shut down. It was lodged under an overhang, a scant 6 feet from a clear shot to the surface. Being only slightly positively buoyant, Marcus was able to force the robot down a few feet and out from under the overhang where it was free to make a slow, unaided ascent to the surface. Surface team members were not expecting Marcus to find the robot so quickly. Third glance at La Pilita, no robot. A few minutes later, ROBOT! By 9:30am John Kerr had the robot out of the water and on charge. The batteries had shut off at 46 volts, so it would be at least 6 hours before the robot was ready for its next mission. This gave the team plenty of time to determine what caused the robot to get lost.
The CMU navigation team downloaded the data the robot gathered during its “star mission”. It had performed the ascents, descents, and traverses perfectly. The navigation fault occurred after the robot finished the “star” sequence at 60 feet, as it attempted to ascend to the surface in the middle of the sinkhole. Nathaniel analyzed the sonar firing patterns as the robot drifted off-course. After analyzing and re-analyzing the data, Nathaniel and George hypothesized that since data from several of the sonar transducers had been filtered out, this caused the robot to be pulled towards the wall. David and Dom joined the discussion. Dom confirmed, apparently as the robot drifted off-course, it was only being guided by one sonar transducer. That was a glaring error and a (relatively) easy fix.
Nathaniel Fairfield explains to the rest of the CMU team what happened on the previous night’s stressful mission. The bot made it almost out, but had a small glitch causing it to get wedged under a rock ledge 18 meters below the surface. The glitch has now been fixed.
pre-morning dive:
morning dive:
afternoon charge session:
evening dive:
With fully charged batteries, morning operations began to attempt a fully autonomous sampling mission. Marcus Gary and Antonio Fregoso provided SCUBA support and video documentation of the sampling sequence. The robot was programmed to replicate yesterday’s mission: descend to 15 meters, acquire an optimal sample target, extend the sampling arm, thrust forward for 30 seconds to increase contact with the sample target, then fire the sample coring device 20 seconds into the thrust forward. The programming filter that compensates for the lack of appropriate ADVM readings was increased to account for up to a 60 second loss of data. Success! Divers observed a perfect sampling sequence.
Divers Antonio Fregoso (left) and Marcus Gary (right) observe a perfect sampling sequence at 15 meters depth.
The robot returned with 2 full bags of water samples and a wall sample. John Spear and Jason Sahl filtered the water samples, cut the filters, and preserved some of the filters in sample bottles with formaldehyde and others were stored in liquid nitrogen for transport.
Left: Earnest Franke retrieves the sample core arm and a water sample bag. Right: John Spear and Jason Sahl filter and preserve water samples.
A second mission ran the robot tethered down to 80 meters to scope out a deep sampling location. Maps of the cenote, which will later allow the robot to select sites autonomously, are not of high-enough resolution yet. A short 20-minute tethered dive down to 80 meters allowed the team to find a sample target. The autonomous run down to 80 meters went successfully. The robot retrieved 1 successful water sample, and during the second water sample pull, the tube clogged. The wall sample successfully triggered and the sample arm got stuck, the robot compensated by first trying to thrust backwards gently then by spinning. The science probe’s camera captured the sample sequence and navigation compensation. Once the robot was docked for charging, the biologists pulled the water and wall samples, processed them and called the mission a success.
During the charging break, Earnest and Ian cleared the clogged intake hose on the water sampling system. They confirmed that all hoses could pull samples. Currently, the team tethers the robot after each mission to download data from that mission. Kantor worked on re-rigging the wireless antenna so that the navigation team can communicate directly with the robot while it’s on the surface. Nathaniel poured through the Doppler velocity log data and the DVL patch code to investigate how the robot deals with velocity spikes. Dom went through navigation code to further refine operation sequences.
Earnest Franke cleans out the sampling intake for the water sampling system.
John Spear (not shown in photo), Jason Sahl, John Kerr, and Ian Meizen (hand on right side) discuss the operation of the water sample collection system. Inset on lower left shows the diagram that John Kerr is drawing.
Once batteries were charged, the robot was lowered back into the water. The team waited 30 minutes for the sensors to equilibrate before aligning the IMU. The robot was launched at 7:15pm on an approximate 4 hour autonomous mission. The robot was programmed to cruise in a star pattern (also known as an inverse negative-hole donut pattern) at two depths. This particular pattern optimizes sonar coverage of the cenote walls to augment point density for the production of a high resolution map of the cenote. The mission was estimated to take approximately 4 hours.
DEPTHX prepares to collect a biologic solid core sample from the shallow walls of La Pilita.
At 1:30am, David Wettergreen woke Marcus Gary with “The robot didn’t come back. It’s an hour and a half late.” The team hypothesized what could have gone wrong, and the fact that soon it would lose all battery power, lights would shut off, and only a faint red strobe mounted on top of the robot would last until morning. Marcus Gary planned to wake up early and search at daybreak.
Left: Dom Jonak in the navigation mission control tent with the day’s outline. Right: The Hydrotech used to verify salinity concentrations during the water sampling cleaning protocol tests.
morning dive:
afternoon charge session:
evening dive:
Objectives this morning focused on completing a successful and repeatable sampling routine while the robot was on tether in order to prepare for a similar unteathered mission. Operations started with simply driving to depth, and initiating the sampling sequence near the wall. CMU worked on coordinating the timing for the sampling procedure, so the robot’s thrusters, sample arm extension, and coring device firing were working together. Several sampling attempts using the close-range sonar transducer integrated with the science package were made to trigger the coring device. The navigation team found that a timed coring device firing strategy produced better results. Once the sampling arm is extended, the robot thrusts forward for 30 seconds to fully contact the wall. 20 seconds into the forward thrust, the coring device fires. After the sampling sequence completes, thrusters reverse and navigation to the point of origin takes over. Establishing that the robot can successfully completing this sequence on-tether will allow the team to test off-tether operations against a known sequence.
Nathaniel Fairfield and David Wettergreen check the sonar firing sequence.
During the battery charging break mid-afternoon, the team focused on calibration of the water collection system and sensing probes. Jason Sahl, John Spear, Earnest Franke, and Ian Meinzen established a cleaning protocol for the sample bag collection system. They tested the cleaning protocol for cross-contamination using 5 saline solutions with varying concentrations. The team took samples of the prepared solutions and Marcus Gary verified the specific conductance of the solutions before and after sampling using a secondary Hydrotech water chemistry sensor. There was less than a 3% difference in original water and the sampled water. The sampling protocol was determined to limit the possibility of cross-contamination. Nathaniel Fairfield and David Wettergreen checked the sonar firing sequence to find that all sonars fire simultaneously.
Marc Airhart manually drives the robot using a wireless joystick to the point of origin before it begins its missions. George Kantor (right) and John Kerr (left) taught Marc to drive.
The evening mission attempted the same task, but this time the robot was untethered. Twice the robot aborted the mission. The CMU navigation team determined that while the robot was thrusting forward, the ADVM was too close to the wall to receive accurate readings which caused the DVL (Doppler velocity logger) to go out longer than the 30 seconds the filter allows. SCUBA divers Marcus Gary and Antonio Fregoso attempted to observe the robot’s performance, but given low visibility and distance thresholds required to not interfere with sonar readings, observation was unsuccessful. The team pulled the robot out of the water for the night and put batteries on charge at 10:30pm.
John Kerr unhooks the bot from lifting slings as DEPTHX gets wet again in La Pilita. The first day back in the water saw positive success in the operations development of DEPTHX.
The team awoke at daybreak and assembled in the Rancho la Azufrosa Palapa (AKA Palapa Internet Cafe) to discuss the plan for the day. George Kantor wrote up the organization schedule for the goals, discussed what needed to get accomplished, and took questions. Most headed to La Pilita by 8:30 to get moving with the tests.
The DEPTHX team meets in the Palapa (left) to go over the daily plan (right), as shown by George Kantor.
Update to come
The group breaks for lunch at La Pilita with the local menu favorite of tacos.
Following lunch, the group began repairs and minor modifications of the science payload. A hydraulic leak was located and fixed, and a rebuilt sulfide sensor was installed on the bot. This work was led by Ernest Franke and Ian Meizen from Southwest Research Institute