Speaker
Description
An essential component of the Smart City concept is traffic, which has a significant impact on the population's well-being and the city's overall economic vitality. The paper aims to propose a new method for Testing and evaluating the Effectiveness of a new Traffic Management System.
The method defined here is the final phase of a complex research project with a general objective of developing an AI-based system for urban traffic management. The primary aim of this research project is to propose a strategy and architecture for an AI-based system that reduces the overall Traffic Congestion Coefficient in any densely populated city.
To achieve the main objective, the research project defined and used a rich architecture, including specialised generators of synthetic data and a society of AI agents.
The Research Methodology consisted of consulting and studying specialized scientific and technical literature on the area of interest, formulating work directions based on each current research stage of the proposed topic, determining and analyzing experimental data, and interpreting and assessing the results. The research results will be published and disseminated to be available to the scientific community.
The objective of the present research was to develop a method that can integrate all the components already defined in the project, test the assembly, and evaluate the results of the leading project.
The advantage of this approach is that it permits the use of multiple scenarios and any combination of trained models to evaluate the results and to highlight the impact of applying different variants of trained agents on the overall Traffic Congestion Coefficient. The feedback obtained after each scenario was used to carefully fine-tune each of the components involved, such as generators of synthetic data, prediction models, and intersection models used inside agents.
The method involves the random generation of a complex network of intersections connected to each other by roads, topologically equivalent to a complex network of roads in a city. For the scope of this research, a reduced set of intersection types and road categories was used.
A visualization component allows for the observation and adjustment of the parameters used in the simulation. The method also provides for the use of statistics collectors at key points to automate as much as possible the final calculations for evaluating the Traffic Congestion Coefficient.
This research presents the promising results obtained in reducing the overall Traffic Congestion Coefficient for several scenarios. A live demo is also available.
