TRAFFIC SPEED PREDICTION USING PROBABILISTIC GRAPHICAL MODELS

The importance of traffic state prediction steadily increases together with growing volume of traffic. The ability to predict traffic speed and density in short to medium horizon is one of the main tasks of every Intelligent Transportation System. Many such systems are currently developed to monitor and control the traffic flow in various states. It is also very important for dynamic route planning applications. Basically, there are two possible approaches to this prediction. The first is to utilize physical properties of the traffic flow to construct a numerical model. This approach is, however, very difficult to implement. Due to the problems with traffic sensor density, it is very difficult to gather enough data to accurately describe the starting and boundary conditions of the model. The other option is to use historical traffic data and relate information and patterns they contain to the current traffic state by the application of some form of statistical or machine learning approach. Authors propose a solution to use a probabilistic graphical models (PGM) for this task. These models are naturally able to capture all complexities in the traffic and incorporate uncertainty of the traffic data. This paper presents an algorithm based on dynamic Bayesian networks (DBN), which are one of the most widely used PGMs for modelling of dynamical systems. Our algorithm was tested on real data coming from the Czech Republic motorways.