Tesla, which disbanded its public relations team in 2021, did not respond to WIRED's request for comment. The camera systems the researchers used in their tests were manufactured by HP, Pelsey, Azdom, Imagebon, and Rexing; None of those companies responded to WIRED's requests for comment.
Although NHTSA acknowledges problems in “some advanced driver assistance systems,” the researchers are clear: They are not sure what this observed emergency lighting effect has to do with Tesla's Autopilot problems. “I don't claim to know why Teslas collide with emergency vehicles,” Nasi says. “I don't know if it's still a vulnerability.”
The researchers' experiments were also solely related to image-based object detection. Many automakers use other sensors, including radar and lidar, to help detect obstacles in the road. A small crop of tech developers — Tesla among them — argue that image-based systems augmented with sophisticated artificial intelligence training could enable not only driver assistance systems, but fully autonomous vehicles. . Last month, Tesla CEO Elon Musk said about the automaker's vision-based system Self-driving cars will be enabled next year,
Indeed, how a system may respond to flashing lights depends on how individual automakers design their automated driving systems. Some people may choose to “tune” their technique to react to things that they are not completely sure are actually obstacles. At the extreme, that option could lead to “false positives”, where a car might brake hard, for example, in response to a child-sized cardboard box. Others may prepare their technique to respond only when it is confident that what it is seeing is an obstacle. At the other extreme, that option could cause the car to fail to brake to avoid a collision with another vehicle because it does not realize it is another vehicle entirely.
BGU and Fujitsu researchers came up with a software fix for the emergency flasher problem. It's called “caracetamol” – a portmanteau of “car” and the painkiller “paracetamol” – it's designed to avoid the “seizure” problem by being specially trained to identify vehicles with emergency flashing lights. Is. Researchers say this improves the accuracy of object detectors.
Arlens Fernandes, an assistant professor of computer science and engineering at the University of California, San Diego, who was not involved in the research, said it appears to be “sound.” “Just as a human being can be temporarily blinded by emergency flashers, a camera operating inside an advanced driver assistance system can be temporarily blinded,” he says.
For researcher Brian Reimer, who studies vehicle automation and safety at the MIT EdgeLab, the paper points to bigger questions about the limits of AI-based driving systems. He says automakers need “repeatable, robust validation” to highlight blind spots such as the sensitivity of emergency lights. He worries that some automakers are “advancing technology faster than they can test it.”