Fourth MSL Landing Site Workshop: Day 3: Engineering constraints
Here's my final pile of notes from the Fourth Mars Science Laboratory Landing Site Community Workshop, taken on the morning of Wednesday, September 29. The powerpoint presentations are now mostly online! I will save you all a lot of reading time and tell you the two big conclusions from nearly everybody's talks:
We can now see just about every hazard that could possibly cause Curiosity to fail during landing using actual data, not models.
There are no hazards that we can see at any of the four landing sites that make the engineers nervous. Any one of the four sites would produce the represent the safest landing site (in terms of predicted hazards) that we have ever had on Mars.
This is astonishing and wonderful. It's also, in a weird way, disappointing. The site selection decision-making would be much easier if one or more of the sites was being eliminated now because it was less safe than the others. But the engineers are happy with any of these sites -- for landing, anyway. I'll write more about the implications of this astonishing fact after I've gone through all these notes. For other posts from this meeting, here is my preview, the Day 1 notes, Day 2 notes, and the official program in Word format, and the presentations. Here goes. One definition: EDL = "Entry, Descent, and Landing," the few very scary minutes between cruise and landed operations, from the top of Mars' atmosphere to wheels down on the ground.
As before, I came in late (I can't drop off my baby at day care so early), so I missed Mike Watkins' introduction and most of Ashwin Vasavada's presentation on the atmosphere and climate at the landing sites. I came in just as he was presenting this cool graph, which Ryan Anderson also had in his blog entry (thanks, Ryan!):
NASA / JPL / Ashwin Vasavada
Modeled temperatures at the four proposed Curiosity landing sites
Modeled temperatures throughout the Martian year at the four landing sites for the MSL Curiosity rover under consideration in September 2010, along with an extreme "test case" (dotted line).
These graphs represent models of daytime (top set of curves) and nighttime (bottom set of curves) ground temperatures that Curiosity would experience at each of the proposed landing sites. You can see how Mawrth is in the northern hemisphere and the other three are in the southern hemisphere; Holden and Eberswalde are considerably further south than the others so have more extreme temperatures. Landing will happen at roughly the 300-sol mark, during late southern hemisphere winter, not far from where all the curves intersect. Ashwin said that the rover is being designed (dotted curve above) to handle temperatures from nearly the carbon dioxide freezing temperature CO2 freezing temp to up to the temperature it was outdoors in Monrovia on Monday -- 40 Celsius. All four landing sites fall within the temperature envelope of this "stress case." Holden and Eberswalde get particularly cold during daytime in winter, near but not reaching stress envelope. Mawrth is more average. [Note the stress envelope is drawn for a southern temperature profile -- flip them for a northern-hemisphere temperature profile; Mawrth is well within the stress envelope, no problems.]
Next up was Ken Herkenhoff, presenting on behalf of Randy Kirk on meter-scale topography at the landing sites as determined from HiRISE digital elevation models. As a reminder, the landing ellipses are 25 kilometers wide, and individual HiRISE images are 6 kilometers wide, so it takes roughly five individual HiRISE swaths to cover the full width of each landing site. (In practice, it takes a couple more than that.) To make digital elevation models using stereo imaging, you need all of those images twice, from two different angles.
Ken said: the terms DEM and DTM are interchangeable. The expected vertical precision is a function of the pixel scale and the base-to-height ratio. Rule of thumb is rho (vertical precision) is = 0.2 pixels. Horizontal resolution is no better than 3-5 pixels -- this is the size of the box you need to correlate images. We have a new algorithm for calculating DTMs that results in higher DTM resolution from same images, reduced noise, but the DTM often has a blocky appearance. This will result in spuriously high slopes. So use combination of new algorithm and old algorithm. One key use of DTMs is calculating slopes on 2-meter length scale [that is, the length of Curiosity's wheel base -- it'd be hazardous to land on a high slope]. There are lots of artifacts at 2-meter slope scale due to MRO jitter. But they can use HiRISE's overlapping CCD design to counter this problem [See here for an explanation of how HiRISE takes images and how the CCDs overlap.] The way they overlap in the hardware can be used to sample jitter -- this is a really tricky solution to the problem! Can reduce jitter to much less than a pixel. Ken showed that the jitter-reduction algorithm produced an impressive cleanup of Nili Fossae DEM.
They have now generated 15 DTMs in the currently considered landing sites and 2 DTMs in deselected sites ("Mawrth 4" and Nili Fossae). The total data volume in these DTMs is 7 times the pre-MOLA Viking topography database for all of Mars. Eberswalde is roughest; Gale and Holden are smoothest, similar to the floor of Gusev; Opportunity and Phoenix are the smoothest. But even for the roughest-topography proposed Curiosity landing sites, acceptable rover traverse paths exist, so all are OK topographically.
Next up was Matt Golombek on rocks at the MSL landing sites. In the past, Matt's talks were always about predicted rock abundance, but you can actually count the rocks. "A definition of a 'rock' from an engineer's point of view: a rock is something that you don't want to land on, and that you don't want to get in the way of your traverse." He described the recent experience from the Phoenix mission. Viking successfully predicted Mars Pathfinder rock abundance. Viking rock abundance predictions were used for Phoenix. But HiRISE changed everything. Three boxes had been selected for Phoenix landing. First picture shocked everyone [with its rockiness]. They had to select a new site. [I wrote about this all when it happened.]
Golombek went on to describe his new automated rock-counting algorithm for HiRISE images. "Each large rock carries a big shadow behind it. No question that large rocks are casting shadows, and if you can measure shadows you can get at rock size. Automated rock count works well until you get down to several pixel diameters; there's a resolution roll-off near 1.5 meters." He mentioned how the algorithm thinks existing Mars landers are "rocks" and gets their sizes correct to within one HiRISE pixel. The algorithm is: "You fit a circle to the terminator, use that to calculate diameter. We have used this to map ten million rocks in 15 square kilometers of northern plains for Phoenix. Distributions match model curves." It also matches the hand-count done by one of Ray Arvidson's students near the Phoenix lander.
One source of error is things that cause shadows that aren't rocks. An interactive graphical tool in their software allows grad students to classify non-rocks. Scarps for instance. Eberswalde is worst case. So we created an interative editing tool. Most of the non-rocks are greater than 2.25 m in diameter. Resolution roll-off is 1.5 meters. Look at the in-between size range and got good matches to model.
99% of the landing risk to the rover is in rocks you can actually see on HiRISE images. Once you get above 6-8% rock abundance you get to 1% chance of hitting a bad rock on landing. The actual rock abundances at all these landing sites? He showed rainbow-colored images where blue was no rocks, and green, yellow, and red indicated increasing rocks. Holden was totally blue: "I think there's one rock in Holden right there." Gale looks rockier, "pretty good match for Gusev crater plains". One bad spot in Eberswalde, otherwise clear. Holden is interesting because TES rock abundances from thermal differencing techniques produce up to 40% rock abundance in Holden -- "I think this is due to outcrop." The final rock abundances as measured from HiRISE: Holden 0.014%, Mawrth 0.015%, Gale 0.047%, Eberswalde 0.054%. Even though there are a few rocky locations, the risk of hitting a rock at any site is very small.
Time for questions from the audience. First one was: there is some talk of ellipses shrinking, which means they could get pushed closer to the "go-to" steeper areas outside the ellipses. Are rocky areas closer to go-to-sites? Golombek: answer depends on site. At Eberswalde, if the ellipse shrinks, it's obvious where to put it -- shift center east, which is away from go to site -- but goal is to land safely. In Mawrth, ellipse would shift southeast, right on top of compositional boundary.
Next question: why are you eliminating the non-rock scarps -- aren't those also dangerous to the rover? Golombek replied that that was a good question, but that those scarps should be noticed by the DTMs and aren't part of the rock abundance analysis. Next question: We are touching down on the wheels. Your analysis is concerned about belly pan rocks. What's the risk of wheel deformation if it landed on a softball-size rock? Golombek: This has been looked at in excruciating detail. Major concern is stress propagation up through wheel into rocker bogies and chassis. In all cases -- has to do with final velocity -- effectively stresses are much lower than was originally thought. There's no issue of breaking any part of rover by having a wheel come down on a rock. The major risk is high-centering on something that hits belly pan and produces damage, or rock on slope that produces higher effective slope.
Next up was Robin Fergason, showing THEMIS maps of landing sites. THEMIS data provides measurements of thermal inertia, which gets at the dustiness of the landing sites. Fergason: "I'll not keep you in suspense. Thermal inertia is not going to be a factor for driveability." Maybe it was a mistake for her to open with her conclusion, because my mind wandered for a bit (first coffee must have been wearing off). Everything seems to be well-indurated or scoured-clean bedrock. Even where there is some dust at some of these sites, thermal inertia indicates it's a very thin layer over bedrock and hence no issue for driving. Everywhere, thermal inertias are high.
Someone asked, can you tell if stuff is indurated material over a soft substrate, like what Spirit got bogged down in? Ray Arvidson answered: We drove over materials in Columbia Hills that looked innocuous until we drove over them. The rover wheels sink. You get slip sinkage and an unpleasant situation. You won't know until you drive over this stuff. On Opportunity trouble we got into was driving into soft stuff. We learned not to drive uphill on soft stuff. On Spirit, you don't know until you drive into it. But Scamander isn't dust [actually Arvidson used the highly technical term "foo-foo dust"]. It's poorly sorted sand. When wheels try to maintain constant velocity, you get increased torque and increased slippage. It's not dust, it's sand.
Next up was Golombek again, giving kind of an overview summary of all the orbital data being input into the landing safety analysis. He began by saying that they have just now recalculated the shapes of the landing ellipses; everybody's maps have been using ellipses calculated from an entry angle as resulting from a 2009 launch. The new ellipses based on 2011 entry have a slight azimuth change -- in fact they have no azimuth at all, they are east-west. 25 km east-west, 20 km north-south.
"These are the best characterized sites in Mars exploration history." He looked again at the thermal inertia data presented by Robin Fergason. He said they made maps of low, medium, high thermal inertia and then we look at HiRISE images and ask, 'can we explain high thermal inertias with a mix of outcrop and indurated material?' Need to avoid dust: "You land in foo-foo dust, you disappear, you're never heard from again." Most places we have already been on Mars, he described as "duricrusty." "What's great about Opportunity is it's a relatively dust-free site. Holden and Eberswalde have not much dust. Gale has higher albedo, some concern, but areas of interest outside ellipse are much lower and so I think they're dust free. A lot of the landing sites under consideration look [in thermophysical and long-baseline slope terms] like Columbia Hills. They all have high thermal inertia -- higher than all since Pathfinder -- "I think this is not a cause for concern, it's a cause for happiness." e went on to analyze large potential hazards -- things like really big craters or dune fields that Curiosity might land in safely but then not be able to roll out of. "There are no large inescapable hazards at any of the landing sites. 39 potential ones assessed. Also no inescapable aeolian bedforms found at any of the sites.
In conclusion: we have complete HiRISE, CTX, CRISM coverage of all four sites. Rock maps directly image all hazardous rocks. We can directly address all engineering constraints with data. There are virtually no inescapable hazards, no unusual radar properties, they are all generally dust free; Gale is a little bit dustier but once you drive out you're OK. The four sites are generally rougher at all length scales compared to previous landing sites, particularly Eberswalde and Mawrth, but not in a way that is hazardous for landing. It looks like this lander is incredibly tolerant to large slopes. Even if you land on a large dune slope, on the slip face at a slope of 30%, you don't drive up that; it's just dumb. You drive out. In every case we saw completely plausible routes. It does not appear that there is any EDL discrimination here.
There is one new concern for landing that the engineers are working. The Phoenix experience was that rockets disturbed sand and dust. Concern here was not that rocket motors were moving sand and dusd ant that wold hurt the rover, but that radar might be picking up on that and measuring an incorrect horizontal velocity. Golombek asked engineer Adam somebody (I forgot his last name) to comment.
Adam: We are concerned about that. We recently concluded about 40 hours of flight data on our radar over Mars-like terrain. There were a couple of instances over larger sand dunes where the helicopter kicked up enough sand to cause change in velocity error associated with motion of sand. We're working that. Atmosphere on Mars is different than on Earth. Radar signature is a strong function of aerial density of sand flow, so we're working through that issue. We're not panicked. We could probably fly through the worst. There are many reasons why we believe Earth observations are boundingly worse than Mars analaog. We believe we would be able to fly through it. It wouldn't be good day for us, but I believe rover would wake up wheels down and unhobbled on surface of Mars. ext up was Devin Kipp on the EDL safety assessment process. We enter Martian atmosphere at a little under 6 km/sec. We spend a large portion of entry flying at a fairly constant altitude, 10 km. We are guiding spacecraft to control our downrange target point. Banking spacecraft. At end of horizontal portion of flight, we decelerate to Mach 2.0 before deploying parachute and slowing to subsonic speed. Then we can deploy heat shield and begin use of radar system. We use this information to make decision on when to take off backshell, altitude waypoints to fly to. Radar-surface interaction continues all the way down until we separate rover from descent stage and begin skycrane maneuver. At that point we stop taking altitude measurements because there is a rover in the way but we do continue to take velocity measurements.
At that point jet cores begin to interact with surface. We could raise dust, do some trenching. The most interesting part of landing is once rover finally gets to surface we have this mechanical interaction where we have this mobility system that is also our landing gear that conforms to rocks and slopes. Hopefully the rover surface interaction perpetuates for the rest of our mission! [a polite chuckle at that.] Nothing about atmospheric interaction is expected to be a discriminator between sites. Everyone looking at images knows that Eberswalde appears more hazardous than the other sites by an order of magnitude. But the difference is at the fourth or fifth decimal place. Since they are all virtually zero, it does not matter.
[I'll note here that it was just so weird for engineers to sound so -- "confident" isn't the word I'm looking for -- I guess "relaxed" or maybe just "not freaked out about the landing." In the next slides Kipp went on to look at some possible failure modes, but he didn't sound nervous about any of them.]
There are a whole lot of things that could go wrong during the landing. We could damage parts, we could tip over if we land on too high a slope. If we landed on a steep surface, we could re-tension bridles and never positively detect touchdown, never separate descent stage, and have problems. Propellant usage could be higher than allocated if we have high slopes and fly more altitude out than we allocated before. Bridles could get entangled with HGA or mast. And we have plumes, which we would never want to directly impinge rover, although there are scenarios where that could be possible. Indirect impingement could sandblast rover, damaging paint. Rover needs to be telecom- and thermal-safe and it has to be able to drive away from its site.
When you analyze all these things, loads and stability are all that matters; they're the ones that need all the work. That's been the focus of a lot of new work. Last time we got together, we were in design-to-requirements mode. We analyzed whether we were safe or not, and didn't have a lot of insight into our capability beyond that. What's changed is that we've moved to a Monte Carlo analysis that stimulates dynamics of touchdown. We feed a set of bounding slope and rock inputs derived directly from 4 landing sites. [They measured the worst-case properties of the landing sites, then added in some padding -- increased slopes, increased rocks, etc. to make the worst case even worse.]
The answer when you roll all this in, is that 99% of the time, we're fine. So this is a very robust rover, very capable of handling whatever we're going to throw at it. We can almost directly see from orbit anything that may kill the rover. The MSL system is robust to the Martian environment. Touchdown failure rates aren't likely to be a major discriminator between sites. They will play into where we place the final ellipse at all four sites. All of the sites are safer than sites we've ever landed at successfully on Mars. Since landing is so safe, they may actually look at reducing traverse risk (risk to mission by requiring long "go-to" traverse after landing) by shrinking landing ellipse to shift it closer to interesting geology, playing EDL risk against traverse risk.
The last talk I listened to was Paolo Belluta, who presented some early work on automated traversability analysis. But it was sort of a work in progress and I didn't take many notes. But this work entered into an interesting conversation I had over lunch. I remarked to one scientist that it's amazing how they are all swimming in data. He agreed but then said that having the data wasn't the same as analyzing the data; he said that much of these data have not been looked at in very much detail.
Looking at the data in detail is one thing the scientists and engineers are going to need to do in order to winnow these four landing sites down. They now need to begin to ask: assume our rover starts in a random location within this landing ellipse. What would the first 100 sols or so of the mission look like? Where we be able to begin addressing the scientific goals of the mission relatively quickly? Will one of these sites result in a year-long wait for valuable science data? Is that an unacceptable risk? What will the traverses look like?
Since that's what the scientists will be doing next, I think it would be fun for you readers to do that next, too. It's not hard to locate the data from all the landing sites. NASA Ames' landing site website has interactive maps: Holden, Gale, Mawrth, and Eberswalde. I suggest you all pick a random point in an ellipse, zoom all the way in, plan your rover's traverse, and see what interesting things you find!