As I mentioned in another post, I have an idea for working train brakes on a model railroad. To be really effective, it would be good to have an engine that can free-wheel and a DCC system that can handle dynamic braking and other such things, but that's another topic..
Here's my day-dreamed plan. I offer it for free to any individual, club, or manufacturer that would like to use it. I would ask that if you make any money at it, I'd love to have a brake-equipped caboose or something. The basic plan is a continuous signal loop along the length of a train. Each piece of rolling stock would have an IR transmitter and receiver on each end - one on either side of the coupler so that between each car there is a TX-RX and RX-TX pair. It would also have an "angle cock" on each end of the car - some sort of switch (latching reed, for example) which would cause the signal to be terminated and returned back up the length of the train instead of being passed through to the next car.
In operation, a locomotive would send a signal (IR on, perhaps) which would be transmitted through each car in the train. At the rear of the train, the signal would be returned up the other side of the loop, back to the locomotive. As long as that signal was present, the brakes would be released and the train could proceed. Having the train separate, leaving a car spotted without a locomotive, or simply turning off the signal (via a DCC function, for example) would apply the brakes.
The mechanical aspect would be simple enough. A length of ni-chrome wire or a solenoid could actuate or release a shoe rubbing against the tread or backs of the wheels. You could make it 100% prototypical, or basic and functional. If the signal and interface was standardized, the means of holding the car would not matter, and anything could be interchanged and work together in a train. Even an un-braked "pass through" car could be used, if needed.
Of course, this design more or less requires a continuous source of power to work. It could be designed so that it requires power to release the brakes, and a spring will otherwise hold them applied (very vaguely like the way the real thing works). That would help, and might allow the system to be automated for a DC layout - apply power to the track, release the brakes - turn power off, stop the train. However, as I envision it, it would really come into its own on a DCC layout. Not only would the continuous power make things simpler, but the ability to control the brakes from the engine would be ideal. Any function (perhaps the one you use to make the brakes squeal?) could be used to turn the system on and off.
One thing that is less than ideal about the system is that it is binary. The brakes are either applied or released. The advantage is simplicity - a handful of transistors and some IR LED's is all you need in each car. A more complex digital system could be designed, having proportional control, but that begins to get complex and costly. Properly designed, a digital proportional system should still be backwards compatible with a a binary system, and it might even be possible to make a binary braked car have proportional brakes in the same way we control motors, with pulse width modulation.
I know this is far to complex to appeal to every modeler. Many of the arguments against command control work against working brakes, and unless some manufacturer or standards organization makes a working system that everybody else can and will use, it will never be more popular than Astrac, CTC-16 or Dynatrol. However, if it could be developed and shared, I think there are a number of model railroaders who would love the opportunity to control the brakes on their trains. Being able to spot a car on a hill without resorting to push pins or brush bristles wold be great. And best of all, I think it would be the first step toward a truly prototypical operation, where train handling, braking, etc. become important, and even necessary.