Dynamic traffic assignment based on drivers´ clustering for the flock traffic navigation model

Abstract

The present document has the objective of dealing with the traffic congestion problem by applying concepts from the Intelligent Systems field. Everyday people face road traffic congestions in big cities causing wastes in time, productivity and accidents. Several techniques from different fields in science and technology have been proposed for road management that deal with different ways of modeling traffic lights, policies, cars and coordination. The Multi-Agent systems field has presented several approaches to the problem, modeling and optimizing different situations. A n approach running over this field is the Flock Traffic Navigation Model, an original traffic coordination method, where vehicles group in "flocks", just like many animal species travel in nature, in order to increase efficiency and security. Given the fact that in most urban areas traffic changes substantially with time, particularly during the rush period, it is important for us to ask how such time-dependent traffic should be assigned to different routes in the road network. The present work deals with the traffic congestion problem, focusing in the assignment of traffic through intersections, considering drivers' activity history. Statistical Analysis and Clustering techniques are implemented based on that information to research new ways of traffic control and joining vehicles into flocks in the Flock Traffic Navigation Model. This is done by means of a Multi-Agent traffic simulation taking into account the reactive nature of the approach.