Today’s transportation system is rapidly growing in scale and complexity, in terms of both the infrastructure and transportation participants. This renders safety and efficiency more challenging to achieve than ever before, and thereby poses unprecedented urgency on the intelligence of the overall transportation system. Fortunately, the increasing deployment of sensing, communications and computing devices at the transportation infrastructure, vehicles, and passengers provide an abundance of data that record the spatial, temporal, environmental and even emotional status of the transportation system and its participants. Such data possess the power of fueling the intelligence in autonomous driving, transportation system modeling, transportation economics, human factors in transportation, and traffic flow modeling and control.

However, due to the personalized, group, social and interactive attributes of traffic scenes, as well as the dynamics of complex spatio-temporal correlation, the data is often distributed and heterogeneous in many aspects and cannot be directly used. This presentation is concerned with the interpretation of data into information and knowledge that could be subsequently exploited to facilitate the intelligence in transportation. In particular, I will discuss tools such as heterogeneous information network (HIN), to describe the meta structure of a network representation, as well as a few case studies including car-following behavior capturing by deep learning, real-time driving style classification based on short-term observations and environment perception of the autonomous vehicles based on multi-vehicle multi-sensor data fusion.