I am very sceptical about a working (and I mean WORKING!) hump yard on a model layout. So far I've seen four hump yards on club layouts, one with the air type and three with bristle type retarders.
Sorry guys - the actions on all of them looked extremely toylike. The cars started like a Formula 1 car at the top of the hump (or not at all). Then at the retarders it looked as if they had run into a wall. Some stopped there, others rolled on, picked up speed again and smashed with a scale 100 mph into the waiting cars down the track.
Ok, I admit, I added a little drama to my report.

But believe me, on every single hump yard the action was jerky, unsafe, random and unprototypical - nothing to admire, rather giving a painful impression...
BUT: This was quite some years back. No computers then! Ray, if you think you can manage air retarders with the appropriate hard- and software, by all means try it. But it will be a major chore. I guess it could be a real research project (BTW: another possible facette of our great hobby). But - it won't be easy!
Like screwysquirrel put it, the main problem is the scaling down of the cars: The laws of physics are against us!

And the second problem is the difference in free-running of the cars. It is extremely difficult to get about the same degree of freewheeling in a whole fleet of freight cars. So you really had to treat each car on the hump individually. Perhaps this is possible with computerized sensors. A real challenge for all computer freaks, ain't it?
Ron
BTW: All four clubs rebuilt their yards to flat switching.
(And thank you Greg for your interesting explanations!)