Problematic Roadway Environments for Automated Vehicles
Terry, Travis N.
MetadataShow full item record
Progress in the area of automated vehicles (AVs) will benefit safety, mobility, and environmental sustainability. Of the roughly 30,000 traffic-related fatalities reported yearly, 94% are due to human error. AVs replace the human element of driving with sensors, cameras, actuators, and algorithms and aim to drastically reduce crashes and associated fatalities. The advent of AVs is also predicted to increase mobility by reducing congestion via ride sharing and more deliberate AV-performed navigation procedures. AVs are also expected to reduce fossil fuel use and improve urban land usage, resulting in positive economic and environmental sustainability effects. AV-related policies and laws are beginning to taking shape, and already differ between some states. While forward-thinking can be beneficial, the consequence of having non-overlapping laws and policies regarding AVs is a potential issue. A complete survey of laws, standards, and regulations regarding AVs for each state is outside of the scope of this project. However, as of the time of this report, 21 states had enacted AV policies, some which impact the testing and regulation of AV technology. Rigorous testing often proves to be a challenge since AV technologies can differ based on a vehicle’s manufacturer, each of whom provides their own suite of systems to perform the autonomous tasks. It is the potential failure of these systems to perform in certain scenarios that impedes progress, both in terms of AV development and policy. This effort seeks to capture, classify, and distribute real-world scenarios that could be problematic for AVs. The developed Visually Confusing Automated Vehicle (VCAV) database of videos, which is available via an online interface, highlights specific elements of the roadway, such as improper lane markings, that AV technology relies upon for guidance. The goal of the VCAV is to promote research in specific areas regarding these scenarios and to inform policymakers, manufacturers, designers, and Departments of Transportation about problematic scenarios. At the time of this report, the database consisted of approximately 12 hours of recorded scenarios. Most scenarios were recorded on restricted-access highways, such as interstates, which are a likely domain for early deployment of highly automated AVs. Other focus areas include inclement weather and any work zone activity. Because these scenarios are often problematic for human drivers, it is important that AVs outperform their counterparts in these situations. The VCAV can be found online at visualconfusion.vtti.vt.edu and will require user login and registration. Once granted access, users can leverage the database to sort by types of scenarios, such as improper lane markings, types of work zones, and video duration. Researchers will continue to curate and manage the database as time and personnel resources allow. More scenarios will be added and categorized within the database as they become available.