2012-01-23

Opportunity Sol 1066 (Spirit Sol 1087)

We're in what is probably the flattest, most featureless terrain this vehicle has been in, ever. You'd think that would be good news: no obstacles, nothing to hit. The problem is that we have this big honking hole in the ground called Victoria Crater, and we need to stay the heck away from it.

Our basic tools for making sure we stay away from something are visodom and autonav. Visodom helps you by telling you where you are, by comparing "before" and "after" pictures to figure out how far the rover actually went -- as compared to how far it thinks it went because of how many times it spun its wheels. The problem with that is that in featureless terrain, the "before" and "after" pictures look the same. We try to fix that by looking at our own tracks, but those have a nice repeating pattern -- a "picket fence effect" -- that can confuse the software. Looking into the crater at the nifty featureful stuff in there doesn't work either, mostly because it's too far away.

Autonav has its own problems in this terrain. With no features, it can't get good 3-D data about the world, so it treats that as a scary lack of data. Essentially, the flat featureless stuff gets treated like a cliff. And since there is a real cliff that we want it to be scared of, we wouldn't turn that off even if we could.

Normally, for historical reasons as much as anything else, Opportunity relies on visodom. Yesterday's drive stopped due to a VO failure caused in turn by the "picket fence effect" when we looked at our highly repetitive tracks. So what do we do to fix visodom's problems here? We have to make our own tracks more interesting.

So that's the plan. We pick a direction -- a direction more or less straight away from the crater -- and plan a test. Mark has a particular idea he wants to try for the first 5m, "but after that, it's up to you," he says. Gracious SOWG chair Larry Soderblom is on board with it; I don't even need to point out that spending a sol or two to do this checkout could save us a whole bunch of sols down the line.

So here's today's drive. We plan a 25m drive, broken into five segments, each of which tries a different test:



  1. Scuff with both front wheels on each step. As we drive backward, we periodically stop to rotate the right front wheel, then the left front wheel, away from us, in order to push up a chunk of dirt.

  2. Same as above, but we wiggle the left wheel instead of scuffing with it. This creates a different pattern on each side.

  3. "Drunken sailor" -- my favorite, of course, but not only for the name. This test tells the rover that it's not allowed to drive straight toward its goal; it must veer a bit to one side or the other. Since the real-world performance of the vehicle is never perfectly symmetric, this should create an interesting pattern out of little more than randomness. Plus, this version is the easiest to sequence.

  4. Every step, do a small turn in place.

  5. Finally, the "mother of all tests": run a step of test 1, a step of test 2, and so on down the line, until each has been done twice.



One of those oughta work. If they don't, we're screwed.

[Next post: sol 1091 (Opportunity sol 1071), January 27.]





Courtesy NASA/JPL-Caltech. Oh, my gods, is it ever flat here. Flat. Flat. Flat flat flat. Except, you know, for the giant scary hole in the ground. But otherwise, flat.

No comments: