Short-term Travel Demand and Traffic Prediction for Urban Areas: A Combination of Data-driven and Model-based Approaches
Since three decades ago, enormous research efforts have been concentrated on short-term traffic prediction which is recognized as an integral part of most Intelligent Transportation Systems (ITS). Although the exact definition varies significantly among studies, the term “short-term traffic prediction” typically entails the forecasting of traffic parameters of interest (e.g., flow, occupancy, speed and travel time) for a period up to 1 hour ahead. Such type of prediction is imperative because the success of many ITS applications, such as Advanced Travelers Information Systems (ATIS) and Advanced Traffic Management Systems (ATMS), are highly dependent on whether short-term traffic conditions can be accurately forecasted in real time.
One of the limitations of most existing studies on short-term traffic prediction is that their proposed methods are only suitable for applications at a freeway, arterial or corridor level, and forecasting at urban networks turns out to be a complex problem. The difficulty in covering a sufficient part of the road network by sensors, as well as the complex interactions in densely populated urban road networks, are among the most important obstacles faced in short-term traffic forecasting. Moreover, responsive predictive methods are expected to be developed and tested given the requirement of a transition from reactive traffic management to proactive traffic management. The enhancement of the decision making capabilities of traffic management system highly relies on the accuracy of short-term prediction with the effects of non-recurrent traffic conditions considered.
With the increasing availability of high-volume, heterogeneous and real-time data for urban areas, a new opportunity has also been presented to today’s transportation researchers to examine, extend and further improve previous methods which were in most cases developed based on relatively poor information of traffic conditions. Therefore, the main research objective of this doctoral research program is to develop methods that can produce accurate and reliable short-term demand and traffic predictions with the increasing availability of various data sources which can effectively help traffic managers enhance urban mobility.
Ding LuoStart/end date: March 2016 – March 2020
Daily supervisor: Oded Cats
Promotor: Hans van Lint