Nowadays, with the emergence of self-driving autonomous cars and smart cities, problems connected with the automation of complex urban transportation networks are at the highlight of technological development. While self-driving cars are more and more common on the roads, man-driven vehicles are still in majority, and that situation will not change within the foreseeable future. Therefore, the artificial intelligence, controlling the self-driven vehicles, must “learn” the method of cooperation with human drivers, and the smart cities need to “study” the travel habits of the urban population.
To cooperate with the human factor, the self-driving control systems need an approximate mathematical model, which is capable to describe the behaviour of human drivers, participating in the urban traffic. This is the main objective of the so-called “microscopic” traffic models, which gives the state of motion (the position, velocity and acceleration) of individual vehicles as a function of time. One of the most common microscopic models is the so-called IDM (Intelligent Driver Modell) model. The IDM model assumes, that the individual vehicle drivers change the current acceleration a(t) of their own vehicles as a function of their current speed v(t), their desired cruising speed vdes, the bumper-to-bumper distance D(t) to the next car directly ahead (considered as „leader” car) and the velocity vleader (t) of the leader car.
Naturally, the behaviour and reactions of individual vehicles can be different, even in case of the same driver depends on the traffic situation, the mood of the driver etc. This variability of reactions is expressed by the parameters of the used microscopic mathematical model. These model parameters can be calculated by the processing of real measured traffic data.
The distance to the leader car can be measured using an optical method combined with image processing techniques. There is a specific object with the same shape, size and dimensions on each
car: the license plate. An image processing software can be trained to recognize objects with a specific shape (e.g. to recognize license plates). To measure the distance of the license plate, the previous calibration process is needed. During the actual measurement in traffic, the continuous image recording can be provided by a commercial dashcam. The actual distance data can be generated after processing the individual frames belonging to the dashcam video record.