Automatic position detection for cars

As I understand the question, at one specific location on the layout you want to move up to 6 cars realistically through a transfer station.

I would vote for #4 as being the simplest to construct. If you look into cutting the rails at appropriate locations you could get exact placement of up to 8 cars, although the real issue would be the effective length of the power pickup on the locomotive, which for the GP38 is 52 feet (about 72mm ?). Sequencing power along the small sections would not be difficult.

My next choice would be an optical solution. The IR sensor you mention the resistance for sounds like a CdS photocell. These are really only good for sensing changes gross changes in light. You'd be better off using an IR emitter and a phototransistor sensor. Use separate devices projecting across the tracks or a reflective sensor. The blocking sensors would require more software control to avoid false triggers due to hardware on the car (ladders, brakes wheels and such). My choice would be a reflective sensor such as: www.jameco.com/z/OPB704WZ-Optek-Technology-Photointerrupter-Reflective-Phototransistor-4-Pin_1872759.html
The performance and accuracy can be enhanced using a stripe of IR reflective paint or tape on the bottom of the car. A disclaimer in that I used to work with such sensors in manufacturing automation, so I'm very comfortable with them.

To me the resistive wheel sets are best for detecting a block occupancy. This is what I'm doing on my layout. To determine precise location you need some very small, precisely defined track blocks on the order of 25mm. Then there is the rolling stock issues. One truck or two with resistive wheels: one requires all cars to have a defined "front", two trucks and you have an issue of detecting one car, or two halves.

I haven't done anything quite like you describes, but these are my thought.

Mark

I tried the solution with graphite paint on wheels at one truck (33' hoppers by FT) ; I got about 10KOhms, but unfortunately it seems that the cars are too light to ensure a good contact, and above all some VERY SHORT successive insulated sections (about 10mm) would be necessary, which is tedious with Rokuhan tracks. And meanwhile I read the topic about "gunked wheels".

So I tested again the solution with magnets and reed switches (two under the tracks, for debouncing) ; it works fine, there is only the problem with the turnouts : when the train runs over them it performs some "push-pull" movements, and this leads to an ugly "accordion-like" effect, due to the springs in the couplers.

Has anybody another solution ? Rigid coupling bars perhaps (the cars will always be run in the same configuration) ? Or what else ? Meanwhile I exclude the solution with several insulated sections, for various reasons.

Thanks for any additional suggestion ...

Except for the Rokuhan contact "springs" ...

And I think this does not solve the necessity for insulated sections !

Unfortunately there are 2 issues with the Märklin switch : first many people say it causes derailments, second it can not distinguish between the 4 axles of a hopper !
I think RFID can IDENTIFY a car, but not perform a position detection ; I was thinking of another solution : the old grid-dip principle, detecting a LC circuit with an oscillator, principle used in anti-theft systems in stores ... The problem with RF frequencies is that they could make interference with radio receivers in the neighborhood if there is too much power, but I think it could PERHAPS be an intersting way to explore ... But 4 are needed (2 tracks, 2 per track for debouncing), so I don't know if it is really useable !

Unfortunately the IR Led and the photodiode are in the same case for a conventional reflection detector ; and so I should need to put the whole detector unit in a kind of rectangular "pit" under the track (painted black inside, and hoping it won't reflect IR, which is not certain) ... and hope that with the increased distance it should still work ; perhaps it is an idea worth to be tested, the depth of the pit would be given by the angle of possible incoming (sun)light. And then I wanted all the way to use a modulated light beam.
Another problem is that I will need to adjust the sensitivity of the detector, to distinguish between the simple dark bottom of the car (resistance tested : about 90KOhm) and a short piece of reflective foil ("produces" about 10KOhm at the detector, so it should be possible). This goes slowly but surely towards a "gaseous plant" ... especially because 4 detectors will be needed (2 tracks, and 2 detectors per track for "debouncing").
I think, I will finally and simply control the movements for this train ... BY HAND !