A probabilistic framework for road traffic reconstruction and prediction based on incomplete data
Résumé
We present some new developments in probabilistic road traffic modeling. The problem at stake is real-time prediction of travel times from floating car data (FCD) coming from probe vehicles. We tackle it using a probabilistic model based on an Ising model, well known in statistical physics, and real-time predictions are computed using the Belief Propagation (BP) algorithm. The Ising model estimation requires only pairwise statistics, which is compatible with the use of FCD data. The behavior of the method is illustrated by a numerical experiment on a space-time highway network.