aerospace and software

I can’t believe NASA would even consider a crewed flight without another uncrewed flight.  The awesomeness of autonomous spacecraft is that you can fly them until you get things right.  My old company has flown its spacecraft numerous times without any major issues, but they’re still doing missions to make sure everything is safe before putting a person on.  Boeing had a MAJOR issue, and what’s troubling to me is that it’s the sort of fault that should have been avoidable with proper ground simulation.  Some things are hard to test on the ground, like getting high confidence in aerodynamics, for example, but proper timing should have possible to ground test.

This also just makes me wonder what is going on with Boeing.  Are they capable of making safe aircraft and spacecraft?  Their recent track record is just not great.  I’d love to know more about their software development process for spacecraft.  I’m familiar with the approach used here in Seattle for aircraft.  It’s industry standard (for aerospace), but I honestly feel it’s flawed, as was obviously demonstrated by the crashes.  I think a lot of the approaches used in software today made more sense when software was far less complex.  The complexity that today’s powerful computers allow calls for rethinking some of the verification approaches.  I think we should be moving away from verification that’s focused on unit testing and moving towards more emphasis on simulation.  Both are obviously done, but my experience has been that as the rigor of verification increases, the emphasis shifts from simulation to unit test, to the detriment of the system’s overall reliability.


3 thoughts on “aerospace and software

  1. becca

    Did you read this article:

    I feel like it rings true based on some of my own experience with Boeing on contracts I worked.

    Also, yes, I am super shocked that this particular issue didn’t arrise in what should have been very simple flight software test. I think we’re all experienced enough to know there’s some nefarious reason that kept it hidden, but seriously. Also, I am so against the rationale of “well if there was crew on board they could have detected the problem (since it was during LOS) and stopped the burn.” Having manual control is great, but that is NOT a successful test. And all the things that could go wrong near a $20b crewed space station? It rattles me. Sometimes working on ISS, you wonder what bad decision is going to write the next accident report, I do feel arrogance there has gotten pretty high after two decades of successful operations; but also, they must be doing something right because two decades no major incidents and safe recovery from lots of close calls. Though the Commercial Crew program is mostly different managers (lots of former Shuttle people), you wonder how that will tamper the arrogance, but I think they have their own problems. I think Commercial Cargo was wildly successful, but the interesting thing about that is its management was (and still is) mostly under the radar – quietly running 2+ space programs for NASA and spending billions of dollars and basically just doing their own thing for the most part. (My bias clearly coming out having worked closely with all three organizations…)

  2. becca

    (Also, with Flight Software, I’ve always wondered how much of the problem is lack of real professionals doing it… I know some amazingly smart flight software people. But almost all of them in “traditional” aerospace companies are aerospace engineers that have decided to allow their careers to turn that way. I find myself wondering where are all the computer professionals – the people who founded the Microsofts and Googles of the world aren’t exactly flocking to traditional aerospace because there is both a pay and a style problem. But at some point having a self trained aerospace engineer, assisted by a second-rate computer scientists, is perhaps not the best way to go with developing code. I think the New Space companies have had a lot more success, with both their style and business models, of recruiting top notch coders than traditional aerospace has.)

  3. admin Post author

    The recruiting of top notch coders has a couple issues. First, software people get recruited and then told to do things the aerospace way. Second, the top managers are not usually retained because salary is insufficient. There are lots of younger coders (though I assume that’s the demographic in general) but fewer experienced senior coders because of compensation.

Comments are closed.