Fast & Furious on a road to nowhere

We’ve all had this feeling. You start thinking about a problem, which appears so trivial. And in that moment you come up with a solution - using technology - which will solve the problem once and for all.

I think that was the excitement that the researchers at the Fraunhofer Institute for Optronics, System Technologies and Image Exploitation, experienced when the German Federal Ministry of Transport and Digital Infrastructure asked them to find new ways to improve traffic flow. They named the project KI4LSA and got to work. The excited researchers then installed high-resolution cameras and radar sensors at a busy traffic-light-controlled intersection in the city of Lemgo.

The next step will be to deploy an AI. Using this technology, the lights will start adaptively changing their pattern and minimize the wait time. According to the researchers, when the system was run as a computer simulation, the results showed an expected savings of 10 to 15 percent.

There is an adjacent project, KI4PED (you can't help but admire how the Germans name projects - short, descriptive, imaginative, memorable) which will also optimize the times for pedestrians to safely cross the intersection. LiDAR sensors [yes, that technology used in (one day soon) self-driving cars] will measure the speed in which the pedestrian crosses the intersection and make sure that the pedestrian is not left in the middle of the crosswalk as the light changes.

Why has nobody thought of this before ...

Here is the thing to consider. You can optimize one intersection and maybe a few more. It is possible to optimize a long stretch on a main road. That has been done and for that you don't need high resolution cameras. You just take your legal speed, adjust the lights to synchronize and start turning them in sequence.

The bigger question is - What are you optimizing for? The highest throughput? The least amount of wait at the intersection, for any particular direction, for any particular day and time? Once you decide what the optimization criteria should be, the questions will start piling up - Should you optimize the lights or the routes (Volkswagen tried using a quantum computer to help with this)? Should it integrate with your Google Maps app on your phone? Should you optimize the driving speed? And how do you optimize the most unpredictable thing in this whole equation - the human driver? All that of course assumes a sunny day with no rain or snow, with few cars around and no accidents.

Once you answer all these questions, you build all the technology, train all your algorithms and are ready to deploy. Now you have to talk to city hall or some other government entity. The director at the sustainability department will be super-excited in talking to you, because this is exactly what the Smart City digital transformation initiative is all about. On the astronomically low chance that this idea would ever leave the desk of that official and move 'higher up', somebody will ask to cost this project out. The number will rival the GDPs of smaller countries. You will realize that it would be far cheaper to buy audio books for every driver with a calming voice talking about how to meditate. It would also be cheaper to build hotdog stands at every intersection to keep hunger away while you are waiting for the next green light. I bet that the Germans haven't thought about that.

Point of all this? Creating a smarter way of controlling traffic is a nice project and technology can help there. The things which are beyond the control of this project are the number of cars on the road, the urban design, or the options people have to take different modes of transportation. Technology can create the illusion that it can solve a problem. Unfortunately, without understanding the real issue, it is just a delusion. And that's the recurrent pattern.