Field Test of Automated Vehicles in Adverse Weather Found Radar and Camera Systems Were Still Susceptible to Rain and Ice, But Not as Consistently as the Vision-only System.

Closed Track Tests Investigated Adverse Weather and Road Weather Conditions Impacts on Automated Vehicles’ Sensors and Perception Systems with Five Test Vehicles.

Date Posted
09/27/2022
TwitterLinkedInFacebook
Identifier
2022-L01149

Automated Vehicles and Adverse Weather: Final Report

Summary Information

Adverse weather conditions can greatly impact vehicle performance and drivers, and their safety and mobility. To make driving safer, automated vehicle technology features may be added to assist drivers with steering and braking.

The objective of this study was to investigate how adverse weather and road weather conditions affect automated vehicle sensors and perception systems and to inform the U.S. Department of Transportation (USDOT) research and development agenda, create awareness among private sector developers and designers on effects of adverse weather on automated vehicle operations. 

METHODS

Five SAE Level-2 (Partial Driving Automation) vehicle models commercially available in the U.S. were tested across two periods on a closed test track located at the Transportation Research Center (TRC) in Ohio, while performing planned maneuvers during repeatable simulated and naturally occurring adverse weather conditions. The tests conducted were developed with the intent to challenge automated vehicle perception systems which contain a combination of cameras (in conjunction with machine vision algorithms) and radar sensors. The first test period, conducted the week of March 12, 2018, exposed selected Level-2 automated vehicles to adverse spring season weather. The second test period, conducted the week of January 28, 2019, focused on adverse winter weather conditions. All test vehicles had the following automated vehicle driver assistance systems: lane departure warning, lane keeping support, lane centering assist and adaptive cruise control.

Lessons Learned

  • Rain and water on the pavement do not disrupt vehicles’ abilities to sense lane lines in most cases. In the vast majority of test runs, vehicles were able to detect the lane line and steer back into the lane, despite the ground being wet.
  • Falling rain water causes vehicles to react improperly, but not for long. Vehicles tested in falling rain often reacted one of two ways. Some vehicles saw the rain as an object and therefore braked, sometimes very harshly. On the other hand, the rain prevented some vehicles from sensing other vehicles on the road and as a result these vehicles sped up. However, in both instances, the vehicles typically realized their mistakes and corrected them shortly.
  • Ice applied to vehicle sensors is a barrier to all AV systems. Every vehicle tested with ice applied to its sensors failed to work properly. Radar and vision-based sensors were blocked and therefore could not function as intended.
  • Attention is needed for slushy snow coverage on radar sensors. Test results showed that none of the  automated vehicles had exhibited difficulty in light falling snow conditions, but adaptive cruise control functions could be impacted with even modest amounts of slushy coverage on parts of a heated radar sensor.
  •  Automated vehicle systems react differently to snow on the ground. Various sensors and systems required an undefined confidence level to remain engaged when snow was on the ground. One vehicle could not sense the lane line even with a very small amount of snow covering the ground.
  • There are gaps for operating automated vehicles in adverse weather. Based on workshops conducted with stakeholders in this study it was found that weather-related limitations of  automated vehicles are not well understood. In addition, there are no clear guidelines as to when to decide whether a trip under automation should begin or continue.
System Engineering Elements

Keywords Taxonomy: