High performance microscopic simulation for traffic forecasting

Accurate modelling of a congested network, where flow breakdown occurs, requires a microscopic model. In addition, it is difficult to model the interface between drivers and an information network without descending to the single-vehicle level, and ensuring that each vehicle makes a set of consistent decisions throughout the length of its simulated trip. Microscopic modelling is computationally intensive, so high performance is a key factor in its effectiveness. If microscopic modelling is accepted for strategic modelling, there is immense advantage in having a single model for all sizes of network, making scalability another key factor. The mass of information produced by a large-scale microscopic model would be virtually intractable without effective visualisation, so that the facts needed to support network management decisions are presented clearly and concisely, without the need for knowledge of arcane data formats. To summarise, it is our opinion that given enough high quality real-time data describing the current state of the network, and an accurate, high-performance, scalable simulation model, short-term forecasting of traffic conditions is possible, given the highly constrained nature of a traffic network.