by pyramids » Sun Apr 03, 2011 8:14 am
The power you need in the lasers can (in principle) be made ridiculously low, because lasers can be made to have extremely low bandwidths and extremely high stability (I did my PhD in an experiment about getting both below 1 Hz, that is a fraction of less than 3 * 10^-15 of the visible spectrum). So it is not like you have to worry at all about someone else's laser pointer or city illumination confusing your sensor --- except for constraints of budget and making similarly involved equipment suitable for operation on a rocket.
Even in practice, however, you can buy filters that allow only one thousands of the visible spectrum to pass through off the shelf, and other equipment, etalons, too, which can further narrow that tremendously. So the only question is focusing enough of your laser light on the rocket, which would work well with a telscope of reasonable aperture, say 8in. Atmospheric distortion is not a problem, because in the reverse direction, it is at most comparable to the aperture based optical resolution limit of such a telescope.
The argument that tracking is more or less all-or-nothing is true. But with retro-reflectors on the rocket, it could be automated, excluding human operator error. We are not aiming for the full 97% reliability as I believe they do in unmanned commercial space flight, and as the track record of Apollo and the Space Shuttle would suggest of having achieved but not greatly exceeded in manned space operations, are we?
Paul, your partition of the problem is formally correct, but the trouble is that the long-term "absolute" accuracy needed from the trajectory-following aspect of the INS implies that it correctly follows all the fast and short-term fluctuations. Because turning first and accelerating later or doing it vice versa leads to different results, you cannot just integrate individual gyro and acceleration measurements to get accuracy on the slow timescale. I think I read somewhere that military grade strap-down INS tend to require <= millisecond time scale measurement update intervals to, well, "struggle" sufficiently well with this issue.