Modeling Vehicular Trafﬁc and Mobility for Vehicular WirelessNetworks
David R. Choffnes Fabi´an E. BustamanteDepartment of Computer ScienceNorthwestern University
Ad-hoc wireless communication among highly dy-namic, mobile nodes in a urban network is a criti-cal capability for a wide range of important applica-tions including automated vehicles, real-time trafﬁcmonitoring, and battleground communication. Whenevaluating application performance through simula-tion, a realistic mobility model for vehicular ad-hocnetworks (VANETs) is critical for accurate results.This technical report discusses the implementationof STRAW, a new mobility model for VANETs inwhich nodes move according to a realistic vehicu-lar trafﬁc model on roads deﬁned by real street mapdata. The challenge is to create a trafﬁc model thataccounts for individual vehicle motion without in-curring signiﬁcant overhead relative to the cost of performing the wireless network simulation. Weidentify essential and optional techniques for model-ing vehicular motion that can be integrated into anywireless network simulator. We then detail choiceswe made in implementing STRAW.
Communication in mobile ad-hoc wireless networks(MANETs) is the focus of extensive research due toits ability to enable distributed applications amongmobile nodes in infrastructureless environments. Ve-hicular ad-hoc networks (VANETs) are a particu-larly challenging class of MANETs characterized bynodes with relatively high mobility (speeds between0 and 20m/s). In addition, unlike many other mo-bile ad-hoc environments where node movement oc-curs in an open ﬁeld (such as conference rooms andcaf ´es), vehicular nodes are constrained to streets of-ten separated by buildings, trees or other obstruc-tions, thereby increasing the average distance be-tween nodes and, in most cases, reducing the over-all signal strength received at each node. Connectiv-ity in this environment is essential for a wide rangeof important applications including real-time trafﬁcmonitoring, battleground communication and othervehicular distributed systems.We argue that a more realistic mobility model withthe appropriate level of detail  for VANETs is crit-ical for accurate network simulation results. Withthis in mind, we designed a new mobility model forVANETs, STRAW (STreet RAndom Waypoint), thatconstrains node movement to streets deﬁned by mapdata for real cities and limits their mobility accordingto vehicular congestion and simpliﬁed trafﬁc controlmechanisms.In a separate paper , we are evaluating and1