If the contact between the wheel and the wire is poor, AND IF there is current flowing hrough that contact, then probably any motion will cause a noise voltage to be developed between the two wires. That noise voltage, thus amplified, can easily trigger a detection that depends on amplified noise. THAT sort of amplifier can be very cheap because noise, frequency response, as well as distortion do not matter.That will be the very cheapest scheme.

 

The actual distance can be measured with a digital millivolt meter Detecting motion would use the noise detector scheme.

 

Yup. Works well with N gauge rails.

A drop or two of Wahl clipper oil rubbed on the rails helps improve the electrical contact with the rails. An old model railroad trick! Works well.

But I really like the wireless mouse approach.

 

How thick are the rails?

 

How thick are the rails?

Whatever thickness makes it work best.

 

If we assume a rectangular rail of 1mm x 2mm, then it will have a resistance of about 0.35Ω/m for stainless steel.
It could be supplied with a 1A constant current source, using a small switched-mode supply, such as one intended for driving LEDs, which would give 0.35V across the length of the rail if using a metre of rail arranged as in post #30.
A precision op-amp with a gain of 10 could drive the input of an A/D on a 3.3V microcontroller (after a bit of filtering to remove any hf ripple from the switcher). A 10 bit A/D would give a theoretical precision of 1mm, but it would depend a lot on how the noise behaves.