I've heard that the FAA has spent over $10 billion on air traffic control software and hardware. I've also heard that their large projects have typically come in very late and and very over-budget.

Yet the U.S. still seems to need a cadre of thousands of Air Traffic Controllers, eyes glued constantly to the screen, making split-second life-and-death decisions about the course of aircraft.

Please, someone who knows more about this, say it ain't so. 

Surely with computers and GPS and radio there is a better and cheaper way to ensure that aircraft don't collide or jam the skies near busy airports.

-- EdPoor

''A little thought about a) the problem domain, and b) the 'real world' of insurance etc will surely convince you that human air-traffic controllers are here for quite a while yet. Solving the scheduling problem is at least NP-hard, if not AI-hard, and even if we could do it, human controllers aren't going to leave much before human pilots do.''

If you want to get a feel for what the human task is like, and think about how it might be automated (or even if it can be automated), you can play around with an ATC simulator. For example, see http://www.xavius.com/. For what it's worth: "split-second life-and-death decisions" aren't required, and procedures for keeping aircraft safely separated are well understood.  The job is stressful, due to the constant attention to detail required, but the white-knuckle near-disaster scenes portrayed in movies are not common in real life.
----
How realistic is the BSD atc game? It's fun, at least.
----
''Canada has recently privatized their air traffic control systems. While it's far from automated, I hear they've managed to streamline a lot of things. Is this true?''

''Of course, some air traffic control problems may also stem from the archaic, low-fidelity half-duplex radio systems that are in use today because FAA bureaucracy has made updating the hardware prohibitively cumbersome.''
-----
One of the hard parts about automating it is that it isn't very automatic on the pilot's side. ATC isn't just scheduling airline
flights; the whole general aviation world is in there too, and for both sets of pilots ATC requests can be all over the map. On
top of that the aircraft under control aren't very automated either -- on a given summer day that Cessna 172 might only be
able to get 500 fpm climb, and everything ATC does ends up being advisory inasmuch as a pilot can always respond "Unable" to a clearance.

You can get a good feel for the ad-hockery in Don Brown's ATC column at AvWeb: http://www.avweb.com/news/columns/182651-1.html

BSD atc ''is'' fun, but it's not very realistic (you're acting as a tower, approach, departure and center controller all at once, you don't have to handoff to other sectors, all of the aircraft are capable of complying with your requirements, there's no
speed control, all the aircraft climb and descend at the same rate, you have to manage their altitude down to the ground, everyone is IFR (so no-one is doing see-and-avoid), you can't tell another sector that someone heading towards yours can't enter it yet... but then again there are no student pilots, no-one busting clearances, no emergencies, no sightseeing flights, and so on.)