It's a good idea but would really require the car to have interaction with the system, or the system to read and interpret individual cars (to assign the token).
I think the advantage with the system shown is they are only measuring throughput rather than keeping tabs on individual cars. Thus I assume it's an easier problem to solve.
Well yes and no. Our proposal wasn't to keep tabs on cars. There's no tracking whatsoever. Lights have no idea of knowing where you have been or who you are: you're just telling them how much pain you've been suffering in the near past. The tokens aren't marked -- they're simply currency so to speak.
The approach does require two-way communication between cars and nearby traffic lights. But the approaches that various people are suggesting, including this article, largely rely on communication as well: just one-way. The lights generally need to be able to detect cars approaching and at quite a distance. This requires the cars announcing their presence: either the lights have a sophisticated sensor system (lots of plates in the road, or a vision system) or -- more likely -- the cars need to have beacons installed which let the lights know they're coming. A great many of these methods also benefit from knowing not only that cars are approaching but where they intend to go -- are they going to turn left or right or straight -- well before they get into their lanes. This requires additional signaling.
What Gabriel Balan picked up on, is that to make the system fairer and more efficient, you need to be able to transfer historical information from light to light. Not only information about how many cars are coming, but what their previous histories were. If you don't want the lights communicating with one another -- essentially tracking your car -- you can do what he did: have the cars build up a history on their own and relay that to the lights as they approach them.
I understand what you're saying. I think in your design you can keep it essentially one-way without needing anything on the car. This could be done by reading the plate and assigning it to the system, which would then allow tokens to be added/subtracted on the system side only. This does, however, mean that all the traffic lights are hooked up together, but still this is much easier than asking that all cars be modified.
The congestion charging system in London works in a very similar way - it doesn't pretend to know when a car is going to come in or out - just reads the plate and stores it in the database.
Yeah, for a lesser investment than that, they could put some basic sensors on so I'm not waiting with 10 other cars while a green light stays on 30 seconds in the other direction for absolutely nobody. "Improving traffic lights" is pretty low-hanging fruit.
I think the advantage with the system shown is they are only measuring throughput rather than keeping tabs on individual cars. Thus I assume it's an easier problem to solve.