2010-10-08

Spirit Sol 627

"We can't say we're stable," Ashitey says, "and we can't say we're unstable." What we know is, we wiggled the wheels, and we slipped. But we slipped more the first and second time, losing about two degrees of overall tilt in the process; the change on the third wiggle was definitely less. But then, at the end, when we cocked the wheels against the direction of slip, we slipped more during that motion.

My take on this is that everything is in fact just fine. We're on a slope; we're tilted; we're going to move when we wiggle the wheels. But we shouldn't be freaking out about that, because it's not what we care about. What we care about is whether we're going to move (or tilt) when we deploy the IDD, which is a different question. And when you look at how little we slip when we wiggle the wheels, I think you'd have a hard time saying we're going to slip when we deploy the IDD.

We're freaking out more because we took the wiggle-cam images -- hazcams before and after wiggling the wheels -- when we first got here, and we saw a change in those images. But the thing is, we've never taken such images before, so we don't really know whether that change is in family with what's happened previously.

At the SOWG meeting, I advocate a staged approach. Take it slowly. On the first sol, just deploy the IDD -- nothing more. Take images before and after this deployment. If we see any vehicle motion, or an unacceptable change in tilt, don't uplink the next IDD sequence. On the second sol, we do just one more thing -- place the MB. Again, accompany this with lots of imaging, and pull the plug on the process if we see any unexpected changes. We'll go on in this way for each of the days until the RPs are back in town (nearly everyone's going out of town for a convention).

This is generally accepted, in principle, although Emily is still reluctant to deploy at all. ("I reserve the right to change my mind," she is careful to say.)

So Ashitey and I move ahead on working out these sequences. They're not hard, but it turns out that this is one of those days with a high debate-to-command ratio. We're only about halfway done when Justin Maki puts together a movie showing each image of the rover sliding as we had it wiggle its wheels several times over the last few days, and then the concerns blow up again.

That isn't the only thing that blows up. Ashitey and Justin get into an argument about whether it's safe to deploy, and Ashitey just loses his cool. Ashitey's trying to make the point that our decision should be based on analysis, not on anyone's gut feeling, but Justin keeps cutting him off. Suddenly, Ashitey's had enough. "Stop interrupting!" he snaps. Next thing I know, Justin's grabbed his coffee and stalked out the door.

This is not a normal sol.

By this time I've already realized two things. One is that I need to stop advocating a position and go back to basics: regardless of any questions about winning or losing, what is objectively the best thing to do?[1] And second, we need to know whether what we're seeing is normal behavior in similar conditions.

We haven't taken wiggle-cams before, which is part of the problem. But we do have very good data on the tilt changes. We always collect this data at 8Hz -- the maximum rate the rover's capable of -- during wheel wiggles. So I've been trolling through past sols, looking for times when our tilt's been high. I find a couple of these, plot the tilt data, and -- yep, looks to me like what we saw is normal. I still have to convince the doubters, but it's good evidence.

But first I have another concern. Justin's a smart guy who's done a lot of good on this mission, and we ought to make sure we've heard him out on this topic. So I track him down -- he's upstairs, in the Opportunity room -- and talk it out with him. In the end, it's just gut feeling on his part (he says so himself), and while I don't want to dismiss gut feelings, evidence counts for more. (While we're having this conversation, Ashitey comes to find me. And apologizes to Justin. So that's back on track.)

Also in the Opportunity sequencing room is Jeff Biesiadecki. He knows the rover's mobility system better than almost anybody, and we don't call him "Paranoid Jeff" for nothing, so we drag him downstairs to have a look through it. Somewhat to our surprise, not that we're complaining, Paranoid Jeff gives our careful, phased plan a thumbs-up. So far, so good.

But Khaled gets wind of this, and objects. Strongly. "I wouldn't deploy in this situation," he says firmly, "no way." He doesn't directly take issue with the historical data showing that what we're seeing seems to be normal, but he raises the Shuttle Fallacy: just because we've gotten away with something dangerous in the past, that doesn't mean we're safe to do it again in the future.

"The problem with that," I point out, "is that in situations like this you don't know whether what you're doing is dangerous, and you're getting away with it; or it's safe, and you're worrying for nothing. You can't tell the difference between the two cases without data, and the data we have shows that the rover doesn't tilt excessively when we deploy the IDD in circumstances as close to this as we've gotten."

The debate rages. Some people are trying to figure out exactly where the left front wheel is, and in particular whether it's solidly planted. Jim Erickson and John Callas get involved. They're arguing with Steve Squyres, who pushes the scientific importance of this particular observation. (Squyres nearly scores an own-goal: when discussing what damage could occur to the IDD in the pose where the MB is deployed, he says, "You know, the steel noseplate on the MB is one of the most robust parts of the arm." You don't say it'll survive the scraping to people who want to hear that it's not going to scrape against anything at all.)

This goes on for, I'm not kidding you, three hours. Emily is stressed. She doesn't want to (seem to) be the one to destroy the IDD, with its irreplaceable science value, but I think that in the end she recognizes she doesn't have a fact-based argument. So she makes the right decision: "We're going to proceed with the deployment," she says at last.

But with careful, objective limits on the tilt changes we'll use to decide whether to pull the plug. (We already have decided that no change in position would be acceptable: even a shift of a millimeter would put a stop to this little excursion. But tilt's another matter: we always see a small change in tilt when we deploy the IDD.) Ashitey and Terry Huntsberger do some additional data mining, to come up with numeric bounds. The final number: 0.3 degrees.

We see more tilt than that, and we switch to an IDD-less, image-heavy backup plan. Less tilt, and we proceed.

Surprisingly, even Khaled eventually agrees to this. And with Paranoid Jeff and Even-More-Paranoid Khaled on board, we must be doing something right.

[Next post: sol 637 (Opportunity sol 616), October 18.]





Courtesy NASA/JPL-Caltech. Slipping slowly down the face of Hillary.




Footnotes:

[1] Always, always, always the right approach. Whenever I've strayed from it, I've regretted it, and whenever I've gone back to that path, I've been glad of it. Always.

No comments: