Automated Vehicles at Highway-Rail Grade Crossings
Based on the 2011 crossing collisions and casualties’ statistics, incidents on highway-rail grade crossings (HRGCs) have been prevalent with approximately 250 fatalities and 950 non-fatal injuries each year in the United States. With an increase in automated vehicle (AV) testing, there is a growing concern on how AVs will interact with HRGCs and whether the technology is capable of traversing current HRGCs. The objectives of this study were to identify what information AVs may need to negotiate HRGCs, to create a concept of operations (ConOps) and Requirements document that explore the base scenarios an AV will encounter when traversing an HRGC, and to develop a set of requirements that addresses the basic functions, characteristics, and requirements of AVs, HRGCs, infrastructure, and trains. An interactive stakeholder workshop was conducted on April 4, 2018 to identify potential safety standards for AV traversals. Select scenarios were discussed from AV experts from Automated Driver Assistance Systems (ADAS), connected vehicles (CV), highway, and railroad sectors. An additional series of 60-minute follow-up meetings was conducted with interested industry stakeholders.
The main takeaways from the workshop’s collaborators and stakeholders are summarized as follows:
- Consider sharing or broadcasting specific train data to AVs approaching an HRGC to prevent potential dangers. Railroad companies may not share this type of information from the train to an AVs approaching an HRGC due to safety and reliability risk concerns. However, AVs need this data, such as train location or arrival times, to traverse an HRGC.
- Complement train data with data from multiple sources. It is suggested that AVs should not depend on a single source of information. AVs connected to the HRGC, train, and other vehicles can complement each other in addition to the AV’s sensor package. Accurate predictions from an AV on potential collisions with detected moving objects are challenging when the object being detected (train) is travelling at a high speed.
- Implement the use of high-definition maps as another input and source of information. Maps are expected to include locations of HRGCs, stop bars, sign locations, and HRGC geometry. Highly automated AVs should not depend on input from vehicle passengers alone.
- Geo-fence areas with known complications to automated driving. The expectations of AV operation should be lowered in complicated traffic networks.
- Install special signage for AVs to detect key information under challenging conditions. Safety risks can be associated with relying on visual inputs from the sensor package alone on the AV, including imposter traffic signs or objects resembling signs that may mislead an AV. Adverse weather conditions will also impact an AV’s ability to traverse an HRGC, and therefore additional special AV signage is needed.
- Add on-board sensors to improve AV performance at HRGCs. HRGCs usually have complex geometry and topography which hinders the ability of an AV to detect visual or audible signals to determine when it can traverse a HRGC. Additional sensors based on different technologies such as Light Detection and LIDAR on-board the AVs will increase detection reliability at HRGCs.
- Send information on varying train lengths to an AV to prevent potential challenges. The accuracy of location, timing, and length of the train will provide a safe way for traversing the HRGC. In fact, the use of connectivity may decrease the dependence on artificial intelligence, when higher confidence-based decisions are available.