Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Vehicles traffic simulation using the social forces model
Submission Author:
Felipe Fauez Wahbe Murad , SP
Co-Authors:
Felipe Fauez Wahbe Murad, Flavius Martins
Presenter: Felipe Fauez Wahbe Murad
doi://10.26678/ABCM.COBEM2021.COB2021-1220
Abstract
This article presents a mathematical-computational procedure aimed at simulating the behavior of passenger vehicle drivers moving in a small urban traffic network, a topic of great relevance for the development and future operation of autonomous vehicles. Traffic network dynamics is a typical problem of Dynamics of Complex Systems, a branch of Dynamics that deals with systems that exhibit unpredictable behavior. Roughly speaking, there are three possible approaches to analyse these category of systems: macroscopic, mesoscopic and microscopic. The macroscopic approach is based on Continuous Mechanics, describing the dynamics of the system by means of partial differential equations similar to the Navier-Stokes ones. In the mesoscopic approach, it is admitted that each element of the system moves in space in a stochastic way, and that its temporal evolution is governed by equations based on the Kinetic Theory of Gases or in lattice Boltzman methods. In the microscopic approach, each element of the system moves autonomously according to predetermined rules built from mathematical criteria. In the literature, microscopic models usually are based either on cellular automata or on rules generated by the Game Theory. However, when the system involves the actions of human agents, models based on social forces acquire great prominence. These models had their genesis in a 1995 article by Helbing and Molnár, who proposed a set of mathematical functions, called social forces, that seek to measure the degree of motivation for the performance of certain actions by human agents who move in clusters. Since then, this approach has been widely used in the simulation of crowd behavior. The mathematical kernel of the system described in the present article is a set of second-order differential equations build upon the social forces model, but, in or case, the social forces describe the motivations of drivers operating their cars on urban roads, thus allowing to simulate characteristic traffic actions such as overtaking vehicles and taking care of obstacles. The results of the simulations indicated that the system is capable of reproducing characteristic urban traffic phenomena, such as the automatic formation of vehicle corridors and the emergence of traffic jams, which allows us to envision a possible future application of this model in autonomous vehicle control systems.
Keywords
social forces, driving behavior, multiagent system
