Solution for traffic jams in cities: traffic lights regulated by genetic algorithms

He traffic light cycle –that is, the determination of the time that green, yellow (or flashing green) and red last in the signal– of a road network can be optimized. The use of computer programs or software based in genetic algorithms would be one of the best options, because with this method you could reduce queues by up to 30%, compared to 20% that are reduced with a conventional algorithm.

The regulation of traffic lights with genetic algorithms would reduce traffic jams by 30%.

“The exponential increase in populations and the way cities are usually organized – for example, the place of work and study far from residence – are some of the factors that cause road congestion”, explains Gabriela Cabrera Lara, Master of Engineering – Infrastructure and Transport Systems of the National University of Colombia (UNAL).

What are the solutions to reduce traffic jams in cities

According to the magister, “the solution to this type of problem does not always imply expanding the offer of road infrastructure. It is possible, through the application of various strategies, such as discouraging vehicle demand, promoting the use of public transport, or the traffic light optimization, with softwareof the road networks that we already have”.

In her master’s thesis, the researcher set out to determine –by comparing a conventional algorithm with a newer one– What is the most efficient method to improve the coordination and operation of a traffic light network?so that cars and pedestrians flow more easily.

“The traffic light cycle is made up of the time the signal is green, the time it is yellow (or flashing green if it is for pedestrians) and the time it is red. TRANSYT-7f, the program we work with, simulates the traffic network and optimizes it by combining and testing different signal times”, explains the magister.

This optimization can be done through a conventional method, such as the iterative algorithm. hill climbingor a more avant-garde method, such as the genetic algorithm.

How a conventional algorithm works and how one of the genetic type does it

He adds that “what the conventional algorithm does is an iterative searchlike this: incrementally changing a single element of the solution and repeating the process until a better solution is found, which is very limited”.

“From elsewhere, the genetic algorithm is capable of combining different traffic light cycles –also called individuals–, crossing them between them and exploring a broader search space in order to find the best combination”.

“This algorithm creates an analogy with the natural selection process described by Darwin, choosing the combination of attributes from ancestor individuals from which better successor individuals can be obtained with the ability to inherit their attributes to the next generations (solutions)”.

How the effectiveness of these algorithms was tested

To put both methods to the test, the magister analyzed a section of Avenida 80 –one of the main thoroughfares in Medellín– from a point at the height of Calle 65 to a point on Calle 75. “To assess the state of the road, we took gauging with drones and a GoPro camera.we measure the volume of cars, maneuvers, delays, queues, all this in order to calibrate the algorithms well”.

Thus, they obtained three scenarios to analyze: the current or real one, the one optimized with a conventional algorithm, and the one optimized with a genetic algorithm, and then the three were compared in the Aimsu Next microsimulator.

What results are obtained with genetic algorithms

“We show that, in terms of efficiency, the genetic algorithm achieves a tail reduction of 30% compared to the conventional method, which achieves a tail reduction of 20%. With respect to delays in the network, it made a reduction of 25% (which is approximately 1 min/km of travel) compared to 15% achieved by the conventional one. In terms of performance in speed, both methods reached an increase of 13%.

This information is useful for decision-makers, since today they are often used software that only program the traffic light cycles, but do not necessarily do optimization searches.

“For the Transit Departments, it is possible to access information on the roads through security cameras, which serves to calibrate, optimize and validate that what is being programmed with traffic lights is efficient,” says the researcher.

“A simple decision, such as implementing a software of a genetic algorithm, would have great impacts on mobility, and even on the environment and public health, since it is known that the less road congestion there is, the less toxic gases are emitted into the air”, concludes the master, who for her research work It was directed by Professor Víctor Valencia Alaix, from the UNAL Department of Civil Engineering.

Deixe um comentário

O seu endereço de e-mail não será publicado. Campos obrigatórios são marcados com *