A Real-Time GPGPU Network Traffic Simulator
Overview:
AutoMatrix is a vehicle traffic simulator that models cars driving through a street map, as well as the algorithms that can be used to route and direct these cars. It successfully simulates vehicle traffic congestion due to heavy usage or traffic incidents, and the effects on surrounding areas. Using NVIDIA CUDA parallel processing technology, the simulator can easily handle millions of cars driving across the map simultaneously.
The goal for the project is to study real-time, dynamic routing algorithms for nodes in sparse graphs, such as street maps. The algorithm will consider current and known future conditions, and will also be able to react to unplanned incidents that occur. In terms of vehicles on the street, this translates to finding the fastest way to direct a large number of cars to get them from their starting points to their destinations.
Screenshots:
 |
 |
 |
 |
 |
Features:
Multiple Car Routing:
Youtube video coming soon...
50_cars_half_hour.avi 50_cars_half_hour.ogv
AutoMatrix can easily route many different cars at once, as seen in this video. Cars from all over the map are all making their way to City Hall using the AutoMatrix grid-based routing.
Hierarchy of Routing Methods:
grid_routing.avi grid_routing.ogv
both_routing_algorithms.avi both_routing_algorithms.ogv
Two different routing algorithms are demonstrated in these videos. The first is a regular A* fastest path routing, considering every street on the map. This gives high accuracy, but can be slow for multiple cars. The second uses an All Pairs Shortest Path algorithm performed on a grid overlayed on the map. This gives up a small amount of accuracy in order to get much faster calculations.
Adaptive Routing:
route_changing_1M_cars.avi route_changing_1M_cars.ogv
In this video, a car from the Northeast is heading towards Center City. By frequently recalculating the fastest path to the destination, the effects of traffic buildup on the routing can be seen visually. Note that there are one million active cars in this scenario.
Traffic Incidents:
weighted_edges_and_crash.avi weighted_edges_and_crash.ogv
There are two things to notice in this video: first, the edges are heavily biased towards City Hall, and the resulting traffic congestion can be seen quickly building up in the middle. Also, a crash occurs on N. Broad St., and traffic backs up around the crash site as well.
crash_and_dissipate.avi crash_and_dissipate.ogv
In the beginning of this video, a crash takes place, and traffic congestion increases in the area. Then, the crash is removed, and the traffic returns to normal.
All videos are available in both AVI and Ogg Video (.ogv) format
Manual:
The AutoMatrix User's Guide is avaiable here.
Papers:
TBA
Contributors:
Daniel Lustig
