This aerial photo shows a simulated car collision scene as seen from a drone.

There were nearly 35,000 fatal automobile accidents on U.S. roads in 2016 (USDOT). In addition to the tragic loss of life, these accidents required lane closures that caused delays, increased congestion, resulted in secondary accidents in the queue, increased risk to public safety officers, and created an economic burden on society. We believe understanding this valuable data can help improve public safety.

Recent work by the North Carolina Department of Transportation (NCDOT) and the North Carolina State Highway Patrol (NCSHP) demonstrated that unmanned aircraft systems (drones) can reduce the time required to clear a collision scene during daylight hours by nearly 77% when used to supplement current collision scene investigation techniques (NCDOT). If these efficiencies can be replicated during low light collision events - about half of fatal accidents occur at night (USDOT) - then integration of drones into the collision scene reconstruction will reduce the total amount of time that lanes are closed, potentially saving nearly $424 million a year in congestion costs (NHTSA) (BLS).

RTI International collaborated with NCDOT, NCSHP, the University of North Carolina Charlotte, and Remote-Intelligence to replicate and extend the 2017 study. We examined the usability of collision data collected with a drone at night over a simulated accident on a closed track on January 20-21, 2018. The team implemented a design with baseline data collected using standard methods and 48 drone trials using three aircraft, four lighting conditions, two flight missions, at three different times (full light, twilight, full dark). The data collected from the flights will be used to create 2d and 3d maps of the accident scene and will be scored based on quantitative and qualitative data quality measures. The reportwill be released by NCDOT later this spring.

The RTI Drone Center has been designing and conducting controlled trials to assess the value of drones for different public safety, first responder, and public health missions since 2012.  We appreciate the opportunity to apply our methods to support NCDOT and NCSHP needs. A special thanks to NCSHP for organizing the test site, creating the collision scene, supporting two days of testing, and showing patience to answer our endless questions. 

Preliminary Results and Lessons Learned

  1. Drones reduce the time required to reopen the road to traffic after fatal accidents (we knew this from the NCDOT 2017 report but it is worth repeating).
  2. Low light data quality is mostly about the lighting solutions...most of the variance in data quality will be explained by what/where/how many lights are used.
  3. Off-the-shelf solutions are nearly ready, but as with any trial we found many small glitches that need to be corrected before they can be considered for use as a routine solution on the road.
  4. Technical expertise is key. Listen to your technical experts.... they know their use cases. We conducted stakeholder focus groups and a series of meetings before we wrote the field protocols.
  5. Keep it simple. We only tested three aircraft and still needed 48 trials to be balanced. If I replicated this study I would limit the number of factors and focus on different lighting combinations.
  6. Documentation is critical. We were fortunate that our staff meticulously entered all required data in the RTI field data management system, including the sensor data and flight conditions, and any surprises about the performance of the aircraft, software, sensors, or lighting systems.
  7. Finally...for the project manager....remember your field methods training, reach an agreement with the team and sponsor on outcome measures before you fly, create a solid design, follow it in the field, develop an analysis plan, and hug your statistician.  The last thing you want is a hard drive full of pretty pictures with no answer to the question “does the drone make it better."

Disclaimer: This piece was written by Joseph D. Eyerman (Senior Research Methodologist, Center for Data Science) to share perspectives on a topic of interest. Expression of opinions within are those of the author or authors.