This article is for the IT specialists working in regional road authorities. Since your recommendations influence the development of transport infrastructure, let us discuss what is happening on our roads in the field of high technology!

This topic has become a real headache. Whenever a region allocates funds for an ITS (Intelligent Transport System), everyone wants to install smart solutions at the busiest intersections in the city center. SpesLab constantly receives requests about this, and we are already tired of explaining that it is a bad idea.

Many years have passed since this Habr article about smart traffic lights was published. During this time, our experience has shown a shocking trend of forcing technology into solving social problems.

Intelligent technologies do not create more road space, they do not compress time, and in many cases they only increase the public image of officials. A smart traffic light works with specific physical parameters that were not used in a static switching system. What does “static” mean? It means the system works without feedback, unlike a dynamic system.

In essence, you could place a person at an intersection, and by observing the traffic situation, they could manually control the traffic lights. There are many factors that can be used to speed up even the busiest intersections — all that is needed is feedback. This feedback is provided by traffic sensors that detect the length of vehicle queues at the entrance to the intersection.

In the past, induction loops were installed under the road surface. Today, smart cameras are enough, which makes the solution simpler and cheaper. (We explain this in such detail because specialists in government organizations change frequently).

In other words, a smart traffic light is an automatic system with feedback. It turns the limitations of passive static settings into a dynamic traffic light switching algorithm.

This was the traditional understanding of a smart traffic light, known as an adaptive traffic light. However, decades of attempts to use such systems have shown their low effectiveness. They work quite well on rural highways, where traffic had to be stopped every five minutes just to allow a single country vehicle per hour to pass.

At night, adaptive traffic lights also work well: we often stand at a red light, even though there is no traffic coming from the other direction. Giving green light only to those who are driving is a great idea, as long as there is no traffic jam.

And this is exactly why adaptive traffic lights have not become widespread in cities. Yes, they can speed up a congested intersection, allowing vehicles to pass through faster, but then they suddenly reach the exit point. Ahead is another intersection without a smart traffic light, and it is not designed to handle the increased traffic load.

We often hear customers say: “The next intersections are far away and there are many side roads, everything will be fine, just make it faster!” We do exactly that, but then there is a tractor in the traffic flow. It accelerates very slowly. While it is speeding up, all the vehicles behind it are also forced to move slowly. We accelerated the intersection, but the entire accelerated flow immediately runs into a traffic jam afterwards. We simply pushed the traffic into a wall — oops.

So what should we do? Ban tractors and trucks from cities? Then we would also have to ban learner drivers, inexperienced drivers who are afraid to accelerate, and, as a joke, all the blondes who take longer to react to a green light.

“What is to be done?” Chernyshevsky once asked and was left without an answer. But we will give you one.

First, we should stop installing adaptive traffic lights. At least because some existing systems occasionally “lose their minds” and give a green light to all directions at the same time — this has happened more than once. Therefore, software should never directly control traffic lights. It should only adjust the phases that are strictly programmed into the controller, where dangerous scenarios such as “green for everyone” are physically impossible.

Second, we need neural-network traffic lights that understand traffic at the level of recognizing each type of vehicle. If a truck is approaching, it should not be placed at the stop line, because the whole stream behind it will accelerate slowly. There are more than 18 parameters that can be analyzed to achieve a significant increase in road capacity. The system can even give priority to buses and other public transport. No dedicated lanes are necessary — the neural traffic light will manage everything itself, including giving priority to emergency vehicles.

But all of this is only the beginning. If there are no additional road lanes or available time gaps, there is only one solution: DO NOT STOP THE TRAFFIC AT ALL!

Sounds bold? This idea was suggested by the neural network itself when smart traffic lights were connected into a single system. Yes, neural networks are designed for self-learning, and this was their conclusion. Before we are called completely crazy, let us explain.

Third, and most importantly: never install smart systems at overloaded intersections without connecting them to neighboring intersections.

The way to relieve a critically congested intersection is to regulate traffic before vehicles reach it — at the previous intersections. By doing this, we can avoid stopping vehicles at the main intersection and instead synchronize their arrival with the green phase. A neural-network traffic light can easily calculate the number of vehicles and their speed from previous intersections, ensuring that a city center area that previously suffered from a complete traffic jam can maintain a continuous flow at 60 km/h or even higher.

Now do you understand what is happening? The number of vehicles does not change — only their speed increases from almost zero to ten times higher. This is the key to fighting traffic jams.

Unfortunately, we often deal with officials who have never even seen basic urban traffic simulation software. A few large monitors on a video wall — which are usually turned off or displaying meaningless information — are somehow called an Intelligent Transport System (ITS) in road transport departments.

Senior officials want to solve big problems, so they immediately demand smart devices only at the busiest intersections in the city center. There is no funding for the outskirts. Even when some funding is available, such an architecture should be built from the outskirts toward the center. The first investments should go into making the peripheral traffic lights intelligent.

In less congested areas, there are free time gaps that can be used to improve traffic flow without any “green waves.” This is especially effective in the direction of leaving the city — it is almost always a winning strategy.

The second stage is to gradually move toward the city center, connecting the peripheral traffic logic into one unified system. Of course, the system will not be perfect — accidents and unexpected situations can disrupt the traffic flow. That is why smart traffic lights should be combined with automatic traffic law enforcement systems, so that drivers have no desire to speed or run red lights.

Either we follow the rules and move quickly, or we choose anarchy — the mother of traffic jams.

Oleg Grishanin
=SpesLab=