About the research
Back of queue crashes in work zones are particularly problematic. In advance of the work zone, drivers are frequently traveling at high speeds, and when they unexpectedly encounter a queue, they have little time for evasive actions, which can lead to a rear-end or run-off-road crash. Although rear-end crashes are usually lower severity crashes in other contexts, within a work zone higher speeds frequently lead to more severe outcomes. Additionally, even at locations within a work zone where drivers should be traveling at lower speeds and be prepared, distraction and a rapidly changing environment can still lead to crashes when drivers encounter the back of a queue.
Some agencies have utilized end of queue warning systems (QWSs), which have real-time sensors located upstream of stopped or slowed traffic to either actually detect back of queues or monitor conditions to predict end of queue locations. QWSs then provide driver notifications of traffic conditions, which ideally lead to lower speeds and drivers being prepared to react to the back of a queue, thus resulting in fewer crashes and conflicts. However, the current generation of QWSs require a very dense deployment of vehicle detectors or sensors within work zones, which can be expensive. Additionally, the effectiveness of some systems can be reduced if prevailing traffic conditions are not accurately captured by system sensors. This may lead either to false alarms that reduce the systems’ credibility or missed detections that can result in an even more dangerous situation than not providing a back of queue warning at all.
The RFP for this project, in general, had suggested an evaluation of different QWSs to assess reliability and impacts on work zone safety leading to improvements and reduced back of queue conflicts. And the RFP specifically suggested the development of a video-based end of queue warning system that leverages recent advances in artificial intelligence to automatic detection and tracking the progression of the tail end of queues in work zones. Some QWSs already employ video-based detection and frequently tie into traffic management centers, which can identify and predict potential problem locations. Most vendors are continually refining their systems and are in a better position to develop this type of technology. Additionally, development of such as system is significantly beyond what can be accomplished within the available resources for a SWZDI project. As a result, the research team will instead seek to improve the understanding of driver behaviors that lead to back of queue conflicts. This information can then be used to improve existing QWSs, which vendors have already spent significant resources in developing.