Recent research has highlighted a significant vulnerability in automated driving systems, particularly their interaction with emergency vehicle lights. A study conducted by researchers from Ben-Gurion University of the Negev and Fujitsu Limited revealed that the flashing lights from police cars and ambulances can lead to what’s termed a "digital epileptic seizure" within image-based automated driving systems. This phenomenon impairs their ability to reliably identify road objects, raising fears that vehicles equipped with these systems could misinterpret emergency lights, increasing the risk of accidents.
The researchers explained that the exposure to these flashing lights causes fluctuations in recognition patterns, particularly under low-light conditions. This can lead automated systems to become uncertain whether an object is a vehicle or not, potentially resulting in dangerous situations near emergency vehicles.
This research was prompted by documented incidents, particularly involving Tesla’s Autopilot feature, where multiple collisions occurred with stationary emergency vehicles between 2018 and 2021. One of the study’s authors noted the complexity of differentiating between various types of vehicles, as emergency vehicles come in different shapes and sizes, which might not contribute to the crash incidents exclusively.
While the findings are concerning, the study did not directly test established autonomous systems like Tesla’s Autopilot. Instead, the researchers utilized off-the-shelf cameras meant for collision detection to understand their vulnerabilities. They acknowledged the uncertainty whether similar vulnerabilities exist in mainstream automotive systems since many car manufacturers employ diverse technologies for object detection that may not be as susceptible to flashing lights.
The U.S. National Highway Traffic Safety Administration (NHTSA) has also recognized that some advanced driver assistance systems may not respond as needed when confronted with emergency lights. This acknowledgment corroborates the researchers’ findings.
As a potential remedy, the researchers developed a software fix dubbed "Caracetamol," designed specifically to enhance the identification of vehicles with emergency lights, which may improve the accuracy of object detection in automated systems.
Experts stress the importance of rigorous testing and validation for automated driving technologies to uncover such vulnerabilities before they lead to accidents. They caution that the rapid advancement of automation should be matched with thorough assessments to ensure safety in everyday driving scenarios.
