Belief Propagation and Bethe approximation for Traffic Prediction

We define and study an inference algorithm based on "belief propagation" (BP) and the Bethe approximation. The idea is to encode into a graph an a priori information composed of correlations or marginal probabilities of variables, and to use a message passing procedure to estimate the actual state from some extra real-time information. This method is originally designed for traffic prediction and is particularly suitable in settings where the only information available is floating car data. We propose a discretized traffic description, based on the Ising model of statistical physics, in order to both reconstruct and predict the traffic in real time. General properties of BP are addressed in this context. In particular, a detailed study of stability is proposed with respect to the a priori data and the graph topology. The behavior of the algorithm is illustrated by numerical studies on a simple traffic toy model. How this approach can be generalized to encode superposition of many traffic patterns is discussed.