Vehicular Networks and Simulation
Cars move on streets, not on random waypoints.
Project Overview
Around the middle of the 2000s, a great deal of vehicular ad hoc network research rested on mobility models in which nodes picked a random destination and moved toward it in a straight line. Cars do not do this. They follow streets, stop at intersections, queue behind one another, and cluster at exactly the places where a wireless protocol’s assumptions are most likely to fail.
If the movement is wrong, the connectivity patterns are wrong, and so is any protocol evaluated on top of them. This project set out to fix the foundation first, and then to check the parts that simulation could not settle.
STRAW
STRAW integrated mobility with traffic: vehicles move over real street maps, obeying traffic control, with car-following behaviour and congestion emerging from the interaction rather than being imposed. The result is a simulation in which vehicle density varies the way it does in a city — thin on arterials between lights, dense at intersections — which is what determines whether two cars are ever in range of one another.
STRAW was written for the JiST/SWANS discrete-event simulator, so it could be used with existing protocol implementations rather than requiring a new platform.
Tooling
Two pieces of infrastructure grew alongside it. SWANS++ extended JiST/SWANS with re-implementations of well-known protocols and mobility models, giving a common basis for comparison. Ceratias added real-time visualization, letting a researcher watch a simulation as it ran, interact with it, and modify it online — and detach and re-attach the viewer so the cost was only paid while someone was watching.
Watching a simulation is underrated as a debugging technique. A mobility bug that survives a summary statistic rarely survives being seen.
Beyond Simulation
The later work in this thread pushed the questions out of the simulator. Exploiting Emergent Behavior for Inter-Vehicle Communication looked at what collective vehicle behaviour makes possible for protocol design, and Down the Block and Around the Corner measured how radio propagation actually behaves in a street grid — where buildings at corners, not distance alone, govern whether two vehicles can hear each other.
Software
All three shipped together as the SWANS++ project on SourceForge, which is
still up — the distribution is straw1_2_1.tar.gz. It is the only one of the
group’s systems from this era that can still be downloaded.
- STRAW — an integrated mobility and traffic model for VANETs, implemented for the JiST/SWANS discrete-event simulator. SourceForge · archived project page
- SWANS++ — extensions to JiST/SWANS, including re-implementations of well-known protocols, mobility models, and a steering and visualization tool. SourceForge · archived project page
- Ceratias — real-time visualization for JiST/SWANS, supporting interaction with and online modification of a running simulation. SourceForge
Our own project pages for these are gone; the links above go to Internet Archive snapshots of them.
Papers
- STRAW: An Integrated Mobility and Traffic Model for VANETs — Northwestern CS technical report, 2005
- An Integrated Mobility and Traffic Model for Vehicular Ad Hoc Networks — VANET 2005
- Exploiting Emergent Behavior for Inter-Vehicle Communication — HotAC 2007
- Down the Block and Around the Corner: The Impact of Radio Propagation on Inter-Vehicle Wireless Communication — ICDCS 2009
People
Lead & PI
- Fabián E. Bustamante (Northwestern University)
Students
Collaborators
- Randall Berry (Northwestern University)