MODELLING OF AN URBAN TRAFFIC SYSTEM USING ARTIFICIAL INTELLIGENCE

Abstract

Attributed to a rise in personal vehicle ownership and the ongoing expansion of cities, the
highest growth of traffic congestion in recent years has been observed. Advances in the use
of artificial intelligence have prompted the question of whether a smart traffic management
system could be developed in order to improve the current state of congestion found in urban
environments. Increases in traffic congestion in recent years has prompted the need for new
and more advanced traffic control solutions. Advances in technology have allowed artificial
intelligence (AI) to have an increasing number of applications. This paper investigates the
use of AI called reinforcement learning in developing a new autonomous traffic control
system based on a realistic traffic system model. This paper provides a critical review of the
relevant surrounding literature. A summary of current technology available in traffic control
was also investigated. Key simulation design decisions are then discussed such as the level
of detail possible on modelling authentic driving behaviour, first required analysing the
different ways in which the driver through the control of the vehicle reacts to their
environment. Subsumption architecture was then identified as an appropriate method for
defining these behaviours. A basic traffic scenario was simulated, and the results show that
reinforcement learning can help in Traffic Management.