The sensors linked to AEB or automatic emergency braking systems may be incapable of detecting high-visibility clothing that is designed to stand out to normal human vision in low-light conditions, according to a study by the Insurance Institute for Highway Safety (IIHS).
Alarmingly, the reflective fabrics and materials that make wearers stand out to human eyes conceal the pedestrian’s presence from AEB sensors per the findings of the researchers.
AEB sensors appeared to detect dummies dressed in black, which are difficult for human eyes to detect, with a high degree of reliability. However, adding reflective patches or strips to the pedestrian test dummies caused the AEB-equipped vehicles to strike them full-speed or slow only marginally in many cases, the IIHS reports.
Predictably, systems from different companies had varying rates of success, though many were quite poor regardless of manufacturer. With no light source except the vehicle’s headlights, a Honda CR-V rammed the pedestrian test dummy 84 percent of the time. A Mazda CX-5 struck the dummy 88 percent of the time.
The CR-V and CX-5 slowed for dummies in ordinary clothing, though they usually still struck it. However, adding reflective strips used on road workers’ uniforms and other high-visibility clothing resulted in the vehicles not even slowing before they hit the dummies. Adding 10 lux of external light somewhat improved performance, but 20 lux of light appeared to make the dummies with reflective clothing invisible to AEB, while those without caused the vehicles to slow substantially.
The Subaru Forester, however, seems to use a different sensor suite, and came to a complete stop in 100 percent of cases both with and without reflective strips, except in one situation. When 10 lux of light was introduced, the Forester struck the pedestrian dummy, but reduced its speed by 82 percent before doing so, dropping from the standard test speed of 25 mph to less than 5 mph.
The president of the IIHS, David Harkey, remarked that “it’s untenable that the clothes that pedestrians, cyclists and roadway workers wear to be safe may make them harder for crash avoidance technology to recognize,” concluding “some automakers need to tweak their pedestrian automatic emergency braking systems.”
The study found that, overall, AEB systems reduce pedestrian crashes by 27 percent in daylight but have almost zero effect at night – which is when most people are struck. Prevalence of AEB systems also varies between automakers. GM has its Automatic Emergency Braking tech equipped as standard in 93 percent of its 2024 model-year vehicles. Cameras or a combination of cameras and radar are used to detect humans in the roadway.
Notably, the NHTSA plans to require AEB in all new passenger vehicles by September 2029. However, with stringent requirements such as systems capable of bringing a vehicle to a complete stop from speeds as high as 62 mph, GM and other automakers have requested the standard to be reconsidered, asserting the requirements are “practically impossible with available technology” and would likely result in significant increases in rear-end crashes.
The fact that automatic emergency braking system sensors may not be ready for such precision is underlined by the current IIHS study. Scientist David Kidd said that reflective strips “probably confounded their sensors,” a fact that IIHS president Harkey described as “a worrisome blind spot.”
Comments
Ah, yes. Trying to legislate scientific advancement, again.
GM is starting to equip some vehicles with a night vision infrared sensor which detects people and animals like deer. If that works for seeing deer on the side of the road wouldn´t it also work for seeing emergency workers and pedestrians at night if its field of view was directly in front?
The AEB is an enhancement for the driver. I presume we still have drivers driving cars. The pedestrians wearing highly reflective clothes or stripes are visible to the driver at night. The driver should be applying the brakes. The AEB is to assist with the pedestrians the driver can’t see.