Introduction

Between 1996 and 2003, a group of researchers from the Informatics Institute (INF) and Mathematics and Statistics Institute  (IME) at Federal University of Goiás (UFG) conducted studies on urban traffic in the city of Goiânia, capital of the state of Goiás, a medium-size city in the center of the Brazil.
 
The goal was to create mathematical models and computer simulations that adequately describe the vehicle behavior on street/road networks in Brazilian cities, since that, according to Professor Bryon Richard Hall's studies during his PhD showed that the softwares available abroad for these matter did not fit some needs in order to realistically represent the way the brazilian traffic behaves. 
 
As a result of these studies, in 2003 this group proposed a software architecture of a decision support system for modeling and analysis of urban traffic, as well a prototype of a computer system built under that architecture. The programming language used for this seminal implementation was FORTRAN WATFIV.
 
This architecture and the prototype used a geographic database, made up of four main sources of information:
 
  • street and avenues data, with their physical characteristics (like length, number of lanes, legal speed limit, among others) describing the structure of the street/road network;
  • information about all possible trips made by vehicles between two different points on the street/road network (i.e., an origin-destination matrix);
  • parameters that describe the functions of latency, used to predict the average time of travel in each street/road;
  • penalty data, which is applied to vehicles  travel time for certain conversions.
 
The four main components of the architecture are depicted below: 
 
Main components of the architecture
 
In the figure, the arrows indicate dependence on the specified object. The architecture is thus divided in order to enable independence of the components, facilitating the maintenance of the software and enabling a component to be replaced without the need for adjustments in the other components.
 
The research carried out by that group was a precursor of a new work started in 2006 a finished in 2008, as a part of Walid Abdala Rfaei Jradi master thesis and under the supervision of professor Hugo Alexandre Dantas do Nascimento. The research described in his dissertation innovates in the previous proposed a software architecture meeting new requirements, such as:
 
  • enable collaborative work and distributed in the analysis and simulation of urban traffic;
  • enable Web access to the system;
  • use map servers (like Google Maps, Bing Maps, among others);
  • adopt a more refined mathematical modeling and simulation algorithms best suited to the problem;
  • migration of the legacy code (in FORTRAN) to a modern language platform, like Java.
 
The implemented simulation uses a macroscopic mathematical approach for estimating vehicle flows in a modelled traffic network, in a way that is most suitable to the Brazilian reality, when compared to models used abroad.
 
Such a system was evaluated by experts who praised the initiative to develop technologies for this area, quite lacking in Brazil. They also acclaimed the importance of releasing the tool as a free and web-based software, allowing collaborative modeling, simulation and analysis of networks by traffic planners. 
 
The developed work also resulted in two scientific publications: one in SOBRAPO and ITSC, both in 2009. 
 
In subsequent years,  the proposed system was used as a framework for supporting new research projects (Master Thesis), focused on aspects not covered by its initial requitements like:
 

 

Based on the good results achieved when applied to many real scenarios, this encouraged the continuation of its development, now under a broader perspective of multimodality, possible in face of the mobility and the data awareness trends of the XXI century.

At the moment, several other features are in an advanced state of implementation and will be soon part of  the system. They include:

  • algorithms for vehicle flow calculation adapted for parallel execution under CPUs – Central Processing Units – and GPUs – Graphics Processing Units. This will allow the increase not only of the size of the networks, but also in the amount of data to be processed (Walid Abdala Rfaei Jradi - PhD Thesis);
  • models and methods for traffic assignment allocation and origin-destination matrix estimation under elastic conditions, with incomplete and/or inaccurate measurements of observed flow (Wanderley de Souza Alencar - PhD Thesis);
  • a lexical analyzer able to dynamically evaluate expressions of cost functions (arc and disutility), so that they can be changed and adapted by the user during the analysis process, without the need of code rewriting and subsequent recompilation (Lucas de Oliveira Pacífico - Undergraduated Final Work);
  • a tool for visualization of informations about O-D travel demands (Halley Weslley Alexandre Silva Gondim - PhD Thesis);
  • a model for the study of improvement for the public transportation (Humberto José Longo and Bryon Richard Hall research project).