Nov 25 – Exceeding expectations!

December 1, 2009

We had an awesome day today. The first mission of the day was to try and get pictures of the grounding line of Taylor glacier. We had attempted this yesterday, but were not able to get the bot close enough to lip of the glacier to take pictures. We still could not get close enough to get pictures of the grounding line, but we were able to take pictures of the lip and confirm that it was ice, and of the lake bottom and confirm that it was sediment. In addition, we saw some interesting low temperatures (less that -4 degree Celsius) during our sweep, indicating fresh water inflow from the glacier. The second mission was an automated bathymetry mission where the robot performed a side sweep of the shore. This mission was entirely automated and worked perfectly. It was very nice to sit back and watch the data flow and the visualizations. I will let the pictures do the rest of the talking.

If you are interested in the details of the visual homing algorithm below, please visit my publications page.

We start every day with a short meeting. Everyone was looking forward to new discoveries.

The bot drove to the glacier melt-hole and surfaced up using visual homing. We added weights to the bot here to enable it to descend and be neutrally buoyant at a depth of about 22 m.

Images taken by the upward looking camera on the bot during visual homing to rise up the glacier melt-hole. During visual homing, the bot moves in a spiral pattern to search for a light source, finds potential targets, identifies a blinking light as the target, centers on it and then rises up the melt-hole. The water was turbid and the collimated light was barely visible unless the bot was right underneath it. The first image shows how the algorithm identified the light but failed to lock on to it since the bot was not directly underneath the light. On the arm of the spiral, the bot was underneath the light and locked on to it. The light appears bigger as the bot ascends the melt-hole.

After ballasting, the bot drove underneath the halocline to the glacier lip. A visualization showing the bot approaching the lip. The beams radiating from the bot are sonars whose returns help us construct this visualization.

The bot underneath the lip of the glacier. This mission kept us on edge all the time because we had to be very careful to not let the bot get wedged in the narrow cavity, yet get enough close enough to capture images.

Image of the glacier lip captured by the forward-looking camera.

Image of the lake bottom beneath the lip showed sediment. This confirmed that the glacier does not go down all the way to the bottom but floats.

Nov 23 – Amazing Taylor Glacier

November 25, 2009

In the current implementation, we spool the profiler down manually using reading from an altimeter. The altimeter tells us the height of the profiler from the bottom of the lake. Altimeter readings are not received continuously – sometimes we do not get readings for many seconds. This is possibly because of the changing reflectivity of the bottom, though the real cause is still a mystery to us.  Hence, spooling down the profiler is partly an art. We need to remember what the starting height was and how much we have spooled. We also used the camera images of the bottom to estimate how far we are from the bottom. Our estimates were a little off at one grid point yesterday, and we spooled the profiler down too much. This caused it to run into the ground. The pump on the profiler ended up pumping a lot of mud through the rest of the mission, making the scientific data useless.

Hence, we repeated much of the mission from yesterday. We cleaned up the dirt from the impeller blades of the profiler pump in the morning.

Bill and John cleaning the dirt from the profiler pump's impeller blades.

We again worked a 14 hour day, but this time we were successful in getting good data. Peter watched the data all day and pointed out some interesting facts. The temperature showed a dip at a depth of 20 m, indicating an inflow of fresh water from the glacier at that depth. Further, the inflow has now reached as far as the bot-house.

Bill made some visualizations of the glacier from the sonar data. No one has ever seen the underwater of the glacier in this detail, and it was very exciting for us to see this. I will let the visualizations speak from themselves.

Top view of the underwater part of Taylor glacier (North up, East right). The light blue cloud at the right is the glacier. The green cloud is the lake bottom.

There is morain on the north side of the glacier. This cross section of the glacier shows the moraine.

The middle part of the glacier has a "lip" or a protruding part at a depth of about 16 m. I posted some images in my previous post. We do not what is beneath the lip. It could be a cave. The grounding line may be beneath the lip. We will explore this in a couple of days by ballasting the bot heavy enough to get below the halocline.

Image taken by upward looking camera on the bot. The lake ice in some parts near the glacier was thin and sunlight penetrated through it. Our melt-hole detector algorithm, that identifies potential light sources for visual homing, had a field day finding so many light sources.

John, Peter and Bill mounted the winch for lifting the bot on the gantry at the new met-hole. John left for McMurdo in the evening.

Bill, John and Peter mounting the winch for the gantry at the glacier melt-hole.

Nov 21 – Here we come Taylor Glacier!

November 22, 2009

We performed a profiling mission on a high resolution grid near Taylor glacier.  The hope is that a detailed analysis of water properties will help reveal the source of fresh water inflow into the lake. We worked a 17 hour day. It was an exhausting but exhilarating day and moved us one step to closer to finding the source.

Below are some pictures from this year. For a complete visualization of the glacier face, see my entry from last year.

A view of Taylor glacier seen in the visualizer. This visualization is continuously updated as the robot moves, using the data from the robot's sonars. The red cloud at the top is the lake ice. The green cloud is the lake bottom. The vertical wall is the glacier face. It looks as if the glacier is floating on water with a 'lip' extending outwards. For visualization of the entire glacier face, see my entry from last year.

An image of the 'lip' of the glacier taken by the forward looking camera on the bot.

A rough day

November 11, 2009

Today was another one of those days when everything goes wrong. We wanted to run a mission towards the east side of West Lake Bonney. We switched to a new 2 km long fiber-optic cable to enable communications for this long mission. Vickie and Bill spent most of the morning spooling out the cable from into a box, and building a strain relief to connect the new cable to the bot. The strain relief reduces the tensile stress on the connector of the fiber-optic cable. I spent the morning organizing data from previous missions and writing scripts to copy, organize and backup data in the future.

We started our mission around 2:00 in the afternoon. After a very efficient profile at the first grid point, we were starting to fell pretty optimistic about the mission. However, in the middle of profiling at the second grid point, about 420 m from the melt-hole, we suddenly lost communications with the bot. The science instrument package (profiler) was spooled down and was hanging about 25 m below the bot. As of this morning, the recovery script to bring the bot home in case of communication failure was still incomplete and we did not have that running.

Fortunately, the bot was resting on the under-surface of the ice and was not executing any motion commands. The only option for recovering the bot was to slowly pull on  the cable. The depth of the lake along the straight line from the bot’s location to the melt-hole was high enough to allow us to pull it back without running the profiler into the ground.  It took us about 3.5 hours to pull the bot back.

The first thing that came to our mind was that the problem had something to do with the new fiber-optic cable. After pulling the bot back, we found that the cable was working fine. Chris found out that he could not ping the router that talks to the main on-board computer. He thinks that the power supply in the main computer pod is flaky and caused the router to shut down. We will debug this problem further tomorrow.


Chris tugs on the fibre-optic cable to pull the bot in as I measure the length of the cable pulled in.


After pulling for about 3.5 hours, we recovered the bot and pulled it out of the melt-hole. The profiler is hanging by the green cable about 25 m below the bot.


We opened up the main computer pod to inspect the router and power supply. Chris and Bart inspecting the pod. The pod-cover is seen hanging from the gantry.

The bot sees things in water!

November 6, 2009

November 5

We had some more difficulties today. While going through the mission start checklist, I found that data from 3 of the 6 thrusters was being received only intermittently. Chris first thought that is was some problem with a driver card in the motor housing and we will have to open the motor pod to fix this. Bart decided to take a look at the screen-shot of the sensor state that we save at the beginning of every mission and realized that one battery had tripped off at night and did not get charged. This all took about an hour to figure out, and we all heaved a sigh of relief when we found that there was no major problem.

We also found that the forward-looking housing had water in it. This was because a fault with the sealing O-ring. Bill, Bart and Vickie were able to fix this without too much difficulty.


Kristof found that the forward-looking camera housing had water in it. Kristof and Bill inspecting the housing.

We finally put the bot down the melt-hole and went on a mission. The objective of this missions was to figure out whether we could use a forward-mounted multi-beam sonar for detecting obstacles. We hung three different obstacles from the melt-hole – an ice-chipper bar, a weight bag suspended from a cable and a PVC pipe. We were able to see all the three objects in the visualizer that was written by people at UIC visualization lab and built-upon by Chris. All in all, it turned out to be a useful day.

Three Objects at 15m 11-05-09

A screenshot of the visualizer that displays the returns from the forward looking vertical multi-beam sonar. The dense red dots at the top are the ice surface of the lake. The objects that we suspended from the melt-hole are clearly visible. Photo courtesy Chris and Bill.


First swim of the robot

November 4, 2009

After many hiccups and delays, we finally got the vehicle to go down the melt-hole into the lake and performed preliminary navigation testing today. We had planned to do this in the morning, but the robot went down about 3 m down the melt-hole and got stuck.  A fish-eye camera showed that the robot was sitting on an ice-ledge in the melt-hole.  We managed to nudge the robot back up. Bill geared up and dived into the melt-hole and used an ice-axe to chip away the ice. Finally, in late evening, we were able to take the robot down the melt-hole into the lake and test our navigation systems. They all worked as expected. We will perform more exhaustive testing and instrument calibration runs tomorrow.

Below are some photos from yesterday.


Bill with a big chunk of ice that we melted after Vickie's input, from what she saw during her dive. Getting the melt-hole right has proven very difficult this year.


We use a fiber-optic cable for communications between our mission-control computers in the bot-house and the robot's onboard computer. This year's exploration missions are longer than last year's. Hence, we bought a new fibre-optic cable. We had some connectors built for the cable by a guy at McMurdo. When the cable reached the bot-house yesterday, we foudn that it would not transmit light from end to the other. This was a big issue because this meant that the communications would not work. After using some diagnostic tool, we finally figures out that one connector was bad. This connector was removed and a new one made. Bart, Kristof, and Chris along with the tech guy from McMurdo.


Second dunk of robot with syntactic foam on. The bouyancy is adjusted manually so that it is neutrally bouyancy by adding/removing small lead weights.


Kristof was using a makeshift wooden log to hook and unhook the bot from the gantry and got a large splinter from it in his hand. Bart helped him get it out.

First dunk of the bot in water

November 2, 2009

We performed the first dunk test of the robot yesterday. This is done by just hanging the bot in water from a crane. The flotation “syntactic foam” is not put on. The aim is to check for leaks in hardware, and make sure the entire system works in water.

Melting the hole took way longer than we expected. The day began by the hole frozen over with ice. Rest of the story in pictures below.



The melt-hole was frozen over with ice.


Jim first broke the ice with an ice-axe.


After breaking the ice, we put the heating coils in water to melt the walls of the hole. Kristof spent almost the entire day tending and moving the coils to get a smooth sided circular melt-hole.


I felt much better and was able to work all day, though still far from 100%.


The bot being dunked into water.

Bart observing the bot. There were no leaks and the profiler worked as expected, except that we could not get images from the sonde camera.


Chris and Kristof chooping off ice for water at the end of the day.