Bio-inspired Neural Network Model Applied to Urban Traffic

Global economic development has the disadvantage of increasing the population in large urban centers that causes an increase of traffic in cities and results in traffic jams. Several types of research have been developed to provide solutions to the problem of urban traffic. This paper aims to implement a Bio-Inspired Neural Network model to control urban traffic semaphores and present the evaluation of the model applied to a real scenario of a big city through simulations. Were considered characteristics such as vehicle speed, streets and avenues lengths, traffic semaphore positions and different traffic demands. The chosen scenario was the Paulista Avenue in the central region of the city of São Paulo, well known for its high traffic demand. The Bio-Inspired Neural Network algorithm was compared with the fixed-time control, which is currently used for the control of semaphore phases in the city of São Paulo. In addition, the behavior of control algorithms was analyzed for low, average and heavy traffic demands. The performance indicators used were the vehicles average travel time and the roads occupation level. The results show that the Bio-Inspired Neural Network model is better in all the simulations made, with the different situations of traffic demands. Tests results show performance up to 44,67 % in terms of average travel time, if compared to the fixed time control.