Traffic detection and estimation
The collection, processing and analysis of transportation data, and drawing inferences from it have been essential parts of transportation research and day-to-day management activities. With the deployment of many operational Intelligent Transportation Systems (ITS), transportation agencies in many large cities now have the capability to collect, process, analyze and disseminate large amount of traffic information, over large-scale networks under their jurisdiction in real-time. For a fully or semi-automated system that manages traffic in a large network, the ability to receive accurate and reliable data from their respective sources and to derive accurate and useful information for decision making becomes more important than in the past. User perceptions on the usefulness and reliability of traffic information, and confidence in the associated traffic management action plans have direct consequences on the funding for future ITS research, development and deployment. It is with this background that this special issue on Traffic Detection and Estimation was conceived. The manuscripts submitted to the Guest Editor were reviewed according to the standard set forth by the Editor-in-Chief of Transportation Research Part C and Elsevier. Unfortunately, due to the page limit, only seven of the highest quality papers could be selected to appear in this issue.
The first three papers share the common theme of processing and interpreting data collected from loop detectors, which is still the predominant traffic sensor in use today. The first paper, by Zhang and Rice, develops linear equations to predict travel times on freeways with dynamic model parameters. Coifman et al., tested a different vehicle sampling and data aggregation approach to estimate the median speed of a freeway segment, using data from single loop detectors, at a station across all lanes. Abdulhai and Tabib present a neural network approach to match inductive loop data over time and space, to identify the same vehicle for anonymous tracking. The fourth paper by Cathey and Dailey presents a different application and approach in traffic estimation. In this paper, the authors have developed a framework that integrates data from an automated vehicle location system and bus route/bus schedule to estimate the arrival times at designated downstream locations. The last three papers focus on the detection of incidents. The authors of these three papers have applied techniques from different fields to incident detection. Teng and Qi present two types of techniques in freeway incident detection: the CUSUM algorithm (from industrial quality control) and wavelet techniques (signal processing). Yuan and Cheu introduce the support vector machine classification technique in incident detection.
Many colleagues in the field have contributed to this special issue.