Wednesday, July 2, 2008

Is this any way to run a robot?


Now I don't mean to speak ill of the very hardworking folks behind the Mars Phoenix lander, but I"ve watched the events of the past few weeks since the landing of Phoenix with an increasing conviction that they don't quite know what they are doing.

Compared with the smoothly methodical -- measured but adventuresome -- progress of Spirit and Odyssey, Phoenix appears a bumbling but lucky buffoon. WIth a sampling system that appeals to require soiling itself, Phoenix is now covered in piles of dirt. Suffering from the mysteriously "out of specification" doors, the critical TEGA looks more like a science fair project that requires the flick of a hopefully-not-noticed finger to operate correctly when the teacher comes buy to issue grades and ribbons.

Now not all of this is necessarily the fault of the engineers and scientists. The soil of mars isn't what was expected. Apparently more couscous than dry flour. The instruments were built, mothballed, and updated over a decade's lifespan.

But such brittle design and superficially haphazard operation causes me to raise an eyebrow. If you're going to an alien world, to an ice pack region that has never before been explored. why make assumptions about the quality of the soil> If you've got years of time to re-test and review your instruments, why should there be unexpected questions? Instead, all the focus was on the EDL systems, the (eventually scrapped) guided re-entry system, the range finding radar, the braking rockets, the potential for interference from the descent imager that eventually cost the project any contextualizing airborne photographs.

This last is a problem I see all too often -- a fixation on the expected or known problem to the detriment of areas unsuspected of posing challenge. Engineers knew that there was a potential for Phoenix to suffer descent phase problems. That's what did in the similar Mars Polar Lander almost a decade ago.

But the other aspects -- at least from my outsider's perspective -- seemed to be taken for granted. TEGA had been built for another mission and was, therefore, regarded as a proven system. The sample delivery system had been developed for another mission and was, therefore, regarded as a proven system. The fact that those other missions never flew or crashed before any operating experience could be gained seemed easy to ignore.

Another classic mistake was made -- though I have yet to hear of any repercussions -- that of modifying existing instruments. Kludging together the proven and the novel seems a simple solution but often reveals itself a short sighted decision compared to the superficially greater challenge of a clean-sheet design. Such was the downfall of the first Ariane V launch -- in which reused software proved incapable of dealing with a novel dynamic environment and the first Delta III launch which fell to a similarly explosive fate for similar goals of false economy.

But part of the problem is just the whole messy nature of the thing. Anyone who knows me recognizes that there is a seed of fussiness that occasionally leads to inappropriate acts of dishwasher loading during social situations. But I'd like to think that this doesn't necessarily carry over to space exploration. No one is there to clear Phoenix up with a portable Dyson or some Clorox wipes. So given this fact, I would expect a design (and operation scheme) that perhaps more carefully guarded the danger of sloppy sample handing leading to contamination or interference.

I look at that sample door and can't help but wonder if, despite the much discussed out-of-tolerance components, it might not have opened a little further without that pile of dirt on top. And there is the microscope...with the same open-funnel-on-top sample collection system that prevents any ability to examine a single scoop of dirt through multiple analytical techniques. And WetChem (a name that always makes me feel vaguely dirty) suffers the same problem of preventing multi-technique analysis, an approach that is the heart of any regular systematic identification of an unknown.

The idea, again, was simplicity. Why have a complex enclosed sample distribution system, "just another thing to break" as my dad would say, when the same arm that is essential for sample collection doubles for sample distribution? But if this simplicity costs data or prevents the full us of the instrument suite, then it has failed.

So we'll see what Phoenix does for us. The data coming back is not without value. The pursuit is worthwhile. But has a false effort at economy or naive quest for simplicity jeopardized the overall return in the manner of CONTOUR, Mars Orbiter, the original design for Dawn, or (lest anyone sense a national bias here) Beagle 2?

No comments: