Syntactic inference for highway traffic analysis

An intelligent transportation system estimates the capacity and emergency response time in highway systems by analyzing the real-time traffic patterns. Due to the stochasticity and the wide spatial spread of the highway traffic, the traffic pattern analysis is an interesting research problem that attracts much attention. In this paper, we develop a novel syntactic pattern recognition model to analyze highway traffic status such as congestion and open flow based on a formal grammar called stochastic context free grammar (SCFG). The corresponding estimator and classifier for traffic status are developed, and we demonstrate that SCFG and its extension Markov modulated SCFG are flexible models for capturing the spatial-temporal traffic patterns. The traffic data is assumed to be collected with a wireless sensor network consists of magneto sensors, and numerical studies are performed to test both the estimator and the classifier. For evaluating the traffic status estimator, real traffic data collected by BHL (Berkeley Highway Laboratory) is used.