Opportunity Sol 919 (Spirit Sol 938)

I take a personal satisfaction in today's activity: trenching with a stuck wheel.

Sure, I'd rather be high-tailing it toward Victoria. In my absence, the team got us to Beagle and beyond. Now nothing stands in our way -- well, nothing except the science team, who feel there's some science work we really need to do first.


The good part is that we're already prepared to do what they want to do. A few weeks before we reached the annulus, I got to thinking about that big old unknown sand sheet, and what we'd want to do once we got there. Trenching was one thing I knew they'd want to do, and I knew we hadn't done it since Opportunity's RF wheel lost the ability to steer, which makes our usual trenching sequence unworkable. So I got Paolo, Jeff Biesiadecki, and (bless his heart) Rob Sullivan to go down to the testbed and work out how to trench all over again.

Squyres said -- very, very politely -- that this was a waste of time. But I had a feeling it would prove otherwise, and it looks like I was right and he was wrong.

Well, it had to happen eventually.

So that makes the sol easy, at least: the trenching routine's already fully worked out, and we just have to send it to the spacecraft.

But you knew it wasn't going to be that simple, right? We're actually finished with the CAM when I realize what's wrong. The new trenching sequence draws a long, shallow scrape with the RF wheel, then turns slightly and makes a second scrape overlapping the first, thus broadening the trench. Now, the turn between the two scrapes is performed in an unusual way. Rather than turning through an explicitly commanded angle, which causes the rover to turn until the IMU -- roughly, her internal compass -- says it's time to stop, the mechanism used in the sequence simply runs the wheels a certain number of revolutions.

When I change this to an explicit turn for simulation purposes, the error is clear: we're turning way, way too far.

I call Paolo and confirm the problem. They worked this sequence out in the testbed, and the number of wheel revolutions to use in the turn was calculated empirically. In the testbed -- in the particular soil material in use there, in Earth gravity, and so on -- this does the right thing. On Mars, it might do the right thing, or it might turn the rover way too far.


The change to the sequence is simple enough; we just have to replace the turn with an explicit one. I'm glad I caught this, but, boy, do I feel dumb for not catching it sooner. My first day back from England, and we're redelivering post-CAM.

Just like the old days, really.

[Next post: sol 942 (Opportunity sol 921), August 27.]

No comments: