There's no indication that the skates would rely on lighting whatsoever to be able to function. Existing cars on the market already ship with sonic- or LIDAR-based cameras that can detect obstructions in complete darkness. Tesla's existing autopilot functionality works just fine on dark unlit roads.
>So then you will need a way for those cars to talk to each other, and typically that means you will need wayside APs to carry that traffic.
You've never heard of a mesh network? Even if AP are needed, it's still considerably less wiring than a full third rail setup and still poses no fire risk.
>And oh, you might want to have calibration points on the tunnel floor so that the skates themselves know where they are, to normalize their position and report back true accurate positioning.
This could literally be something as simple as a handful of RFID tags on the floor of the tunnel. Not costly and not hard to implement.
>You underestimate the amount of engineering work needed to do this safely and proving that it's safe. Are the lifts open lifts where the skates drive on and open up their doors? If so, that "marker" will need to be substantially big to allow for overshoot/undershoot and still be safe to open their doors (think, pinching, trapping, etc).
>If so, the lifts will need to be substantially big to allow for a number of these skates to ascend and pick up/drop off passengers.
You're really overestimating the amount of engineering work needed to be done. Again, my $100 roomba is able to park itself within millimeters onto a dock, and it's never had an issue. Automated subway systems all over Asia also operate with only inches of margin for positioning themselves onto station platforms and aligning their doors, and they don't seem to have an issue.
The concerns you're raising are already solved. The 'magic' of this system isn't in the self-driving vehicles part at all, as that's already being done with automated transport systems the world over. The 'magic' is in the tunnel construction itself.