@ frank ly
Quote : "If you want precise positional information about a distant object, surely you'd need the individual sensors to be as far apart as possible...".
The galaxies are too far away to use trigonometry to measure distance. We can use that for neighbouring stars, but once you get past a certain distance, it becomes very inaccurate.
So this about grabbing the spectrum of the galaxies, which gives us the red-shift, which tells us how far away the galaxy is far more accurately at these distances than trigonometry would.
We've done this already on a smaller scale, so this is about doing it on mass.
Quote: "So why are these fibre optic heads placed so close together that they have to be careful not to make them collide with each other when adjusting their aiming line?"
Don't think aiming, think more filter. It's one fibre per galaxy.
Imagine a disk (i.e. a disk of aluminium about a meter across), now drill holes in that disk that precisely match with the relative positions of galaxies in a specific area of the sky. Stick this plate at the end of a telescope (where the camera normally is), and point it at that section of the sky, so that the light from each galaxy lines up exactly with the holes on your drilled plate.
Now direct via fibre optics the light from each hole to a sensor, and you can measure the spectrum of each galaxy, one fibre being the light from one galaxy.
This was basically the Sloan Digital Sky Survey, check out wikipedia etc for some pics.
This new work is to automate the process, so rather than drilling metal plates, and fitting the fibres, a plate for each section of sky, you just move the fibres around via the 'robots'.