A Scalable Cellular Automata Based Microscopic Traffic Simulation

Simulation, Traffic, Scalable approach, Bitwise operation

Abstract of the same name paper presented on IEEE Intelligent Vehicles Symposium 2011.

With increasing traffic densities and safety requirements, intelligent transportation systems (ITS) become more and more important. Traffic simulations can assist in the area of state prediction, which could be very helpful for ITS. A reasonably good prediction can be obtained using microscopic traffic simulation model. Such models distinguish and trace every single traffic entity. Due to many entities and very complicated dynamic relations between them, microscopic traffic simulation requires considerable computing resources.

On account of their simplicity and suitability for acceleration, cellular automata (CA) based models have become popular in the area of microscopic traffic simulations. In CA model, a piece of road segment is represented by a CA cell. It is possible to perform millions of updates per second and it has been shown they are able to cover all basic phenomena occurring in real traffic flow. On the other hand, they were also criticized for some unrealistic behaviour. Therefore, simple CA based traffic model has been extended and updated to the advanced cellular automata model in order to adapt simulation model to local conditions (based on local measurements) \cite{iv2011}. Compared with traffic fundamental diagrams, our model achieves better precision. One of the most important updates to the model was the elimination of unwanted property --- stopping from maximum to zero vehicle speed in one simulation step. This was accomplished by adjusting cell's local transition function. We are also highly memory efficient in comparison to other CA based traffic simulation models because of effective bit level cell state encoding. Our model was implemented using OpenMP specification. A parallel implementation of the proposed model enabled an almost linear speedup on the quad core machine. As a result, this allowed us to run a simulation multiple in real-time, so the traffic state of very large-scale networks can be precisely predicted, for example, with various scenarios.