This project was carried out to build from its most basic form the elementary calculation unit of neural networks known as perceptron. The implementation involved both hardware and software, as well as the communication between elements that make up the system. The objective was to create a small-scale solution that could be scalable to an everyday situation.
Thus, it was decided to create a system capable of distinguishing between people (in our case plastic human figures) and cars (toy scale), but with enough detail to be able to represent the real shape of these entities. In the project, image pre-processing methods were used to obtain information that was classifiable for a perceptron. Obtaining the classification response emitted by the perceptron, the signal is sent to a microcontroller in order to control a traffic light for vehicles and one for pedestrians.
This emphasizes the importance of data analysis in order to computationally understand the variation of physical quantities in an environment. As well as the capacity of the algorithms belonging to the area of artificial intelligence to issue an appropriate response to these magnitudes. Finally, it is shown how the linking of electronic components allows us to efficiently obtain data that are processed by the perceptron, and helps us to exercise control over a physical system.
Theoretically analyze the problem and put into practice the construction of a complete system, hardware-software, for the creation of an intelligent traffic light using a Rosenblatt perceptron as processing unit.
This project formally evolved into a paper presented at an artificial intelligence congress held in Mexico, the slides of the paper presentation can be found in the project repository.