主题：Average Speed Revisited — from Big Data back to Basic Definitions
Dr. Lee D. Han is a Professor at the University of Tennessee in the Department of Civil & Environmental Engineering.
Dr. Han has a wide range of research interests. He has published widely on the subject of mass evacuation including the monumental super-node methodology that optimizes the destination assignment and route selection simultaneous. He also established the measures of effectiveness (MOEs) for mass evacuation operations. Dr. Han published a series of five papers on algorithms that significantly improve vehicle tracking capabilities in a very challenging environment (US) where there are some 3,000 different plate designs and the matching rate is typically in the 30%, which is subsequently improved to 98% with Dr. Han’s text-mining algorithms. Dr. Han also worked on unsupervised machine learning algorithms publishing successful learning algorithms for the applications of real-time incident detection, weigh-in-motion thresholding, automated vehicle tracking, and end of queue identification. He also published in areas including traffic signal control, transportation policy, traffic safety, driving simulator applications, and other theoretical as well as practical subjects. Dr. Han has mentored 15 doctoral students and over 50 master’s students to fruition and is currently working with seven PhD students, two post-docs, and a team of visiting scholars in his research group.
Dr. Han earned his PhD from U.C. Berkeley and taught as an Assistant Professor at Virginia Tech for two years where he wrote the proposal that was awarded the multi-million-dollar Intelligent Transportation System Research Center of Excellence (ITS RCE) at Virginia Tech. Joining the University of Tennessee (UT) subsequently, Dr. Han also wrote the proposal that won the Southeastern Transportation Center, which has brought in more than $20 million competitive federal dollars since the mid 1990’s. Dr. Han has been jointly appointed Oak Ridge National Laboratory (ORNL) Collaborating Scientist in an agreement between the ORNL and UT. He was also elected the Chair of University Research Council. He currently serves as a full member of TRB’s Managed Lane Committee and Emergency Evacuation Committee. Dr. Han was previously the coordinator of the University’s Transportation Engineering program. He is currently the Director of the Driving Simulation Laboratory and the Virtual Traffic Management Center at the University of Tennessee.
The concept of average speed has long been well established, but in practice it took a long time for practitioners to understand the difference between Space Mean Speed and Time Mean Speed. In the age of Big Data where every smart device could serve as a “probe” to potential provide real-time speed information, how average speed should be calculated correctly can still be confounding to some.
In 2010, United States’ SAFETEA-LU legislation requires that all states to provide real-time traffic and travel conditions information for all Interstate routes by November 2014 and on other major roadways by November 2016. This legislation sent all state departments of transportation (DOTs) scrambling to find means for monitoring and measuring traffic/travel conditions on their myriad highways. While most states had already instrumented various types of traffic sensors along most metropolitan Interstate segments, these sensors are often considered too cost-prohibitive to be deployed in lower class and rural areas. One of the popular technology solutions is based on the ubiquitous smart phones and smart devices with real-time cellular connectivity and GPS location functionality, which is becoming a commercially viable alternative for measuring traffic speed on all highway links. However, the quality of travel time and travel speed information measured by different technologies are often only reported by vendors (such as INRIX and HERE) and not compared on the equal footing again other technologies. In this talk, Dr. Han presents an effort to evaluate technologies including cellular-based probe vehicles, Bluetooth, Remote Traffic Microwave Sensor (RTMS), and Google data with ground truth acquired by Automated License Plate Recognition (ALPR) systems. Furthermore, in an effort to more accurately measure complete traffic stream information, namely average speed, flow, and density, Dr. Han revisits the basic definitions of these traffic stream characteristics to explain the common errors and proper approaches.