Final project in electronic engineering
"Autonomous square system" is a unique system, based on artificial intelligence in deep learning, which combines a traffic circle and a traffic light junction. The system operates autonomously and only starts working when an integration problem arises, and at the rest of the time the square is governed by the rules of the traffic circle. After each treatment it's provides a detailed report describing the condition of the square and the problem of integration, the way of treatment, the condition of the square after the treatment and the degree of success. The system can be installed on existing squares and is therefore characterized by great flexibility that allows the traffic engineer to optimally adapt it to each square specifically through the manager interface.
The system combines two sub-systems:
-
PC system:
Responsible for describing the situation in which the square is at any given moment and locates integration problems by analyzing the information coming from the sensors and cameras scattered in the square. In addition, it fills in the treatment reports and is responsible for coordinating the system to the square itself and its road conditions, when installing the system, through the system administrator interface. The system is written in Python and included about 2,000 lines of code divided into a main program (main) and two libraries (tables_fun and data_transfer). -
NUVOTON system:
The system is installed in the square itself and is responsible for dealing with integration problems through the management of the traffic lights in the square. Finding a unique solution to each problem, testing its effectiveness and finding a new solution if necessary. At the end of the process she also evaluates how successful she is in solving the problem. The system is written in C and included about 700 lines of code divided into a main program (main) and a secondary file (Secondary file).
In traffic circles, there is an integration problem, that arises when there is a dominant route in the traffic circle that blocks an entrance path from being
integrated into it.
Example of an integration problem:
-
I’ve learned how to train neural system to object detection.
The project itself uses trained models for reasons of accuracy and reliability, but during the project I trained a system of neurons to object detection to identify my face and photographed the result (click here). - I designed an algorithm to locate issues and to find a unique solution for the problem (i.e. not from a solution stock).
Model name: SSD MobileNet V2 FPNLite 640x640
Results:
