this post is part of the January posting spree initiative Check it out and you are participating for one of 100 premium membership that will be given away this month
This kind of traffic control is the closest thing to AI traffic control active now. It is reactive to the reality, but it uses sensors and is "blind" to everything else, also, it needs to be programmed previously with the goals to what it's being used for, so it keeps the traffic in a certain pattern as much as possible.
The studies required to see how it will benefit in abnormal conditions have not been performer due to the length this will require. Plus the installation is costly and require attention after, so says This article
The Lidar technology is already used to measure distance, it is also used in a fine way to recognize objects, like vehicles, humans, buildings, etc. It's used on driver-less cars, on rumba robots, there is also of task based works online where people train the programs that will read the information from this sensors, since for a machine to learn it needs way more data input than a human.
The information to train AI systems is already there, AI has been already trained with that information which makes it very feasible to use in the traffic control.
The difference between the Adaptive Traffic Signals and directly using AI, from my perspective, could be in the less need of intervention once the system is up, since it will be able to make adjustments in a wider range of situations even if not pre-programmed, in the basis of it being able to combine a huge amount of data to determinate the actions to perform.
AI could use a camera and combine the vision of multiple cameras to control the traffic, instead of just seeing the pedestrian and cars in front. This way if there is a reckless driver in one lane, the system can make everyone else stop / move out of the way, and save a lot of people from being involved in an accident.
This level of AI usage will require an input of data so huge, it will require too much computer power, which is a problem to be solve not too long in the future with the more and more fast and capable processors being created. The training of the AI, even when the data is already there, will require some more instruction for it to react to unexpected situations.
It could send unnecessary alerts when a driver goes too fast but stops ar the right moment. Which will then, for no reason (although this is the driver's fault for trigger the alert) delay the traffic while it is put in circulation again.
Even when the complications, I think AI being integrated in the traffic control but with added functionality like trigger airbags or spawning road bumps. It would certainly be a smart city... well.. at that point... would it still be needed or the cars would be able to control all of it by themselves?
Designed by @jes.seth
Designed by 
