Tesla Autopilot tricked by painted wall, exposing camera-only limits

Tesla's push for fully self-driving electric cars has run into a wall. Figuratively, and literally. A simple test has fooled its Autopilot system. A YouTuber, Mark Rober, showed how a painted wall could trick a Tesla into crashing, exposing the limitations of its camera-only approach. Although far from being scientific, the test raises a lot of valid questions about the safety and reliability of Tesla's approach to autonomous driving technology.

Tesla, led by its controversial CEO Elon Musk, relies solely on cameras and artificial intelligence for its Autopilot and Full Self-Driving systems. Other manufacturers use additional sensors like LiDAR and radar. A former NASA engineer, Rober, decided to test Tesla's system using a scenario inspired by the classic Looney Tunes cartoon. He built a wall and painted it to blend seamlessly with the surroundings. It looked as if the road carried on - when, in reality, it didn't.

Two vehicles were involved in the test - an older Tesla Model Y and a Lexus SUV equipped with LiDAR technology. Both vehicles initially performed well in a test where a simulated child appeared in front of them. However, when smoke and, later, rain-simulating water were added - things changed quite a bit. Tesla struggled with telling the child apart from the surroundings and, in both scenarios, didn't stop.

At this point, there was no surprise when Tesla was faced with the painted wall. The LiDAR-equipped Lexus detected the wall and had no trouble stopping safely. The Tesla, relying only on its cameras, failed to recognize the obstacle and went full-speed through it.

This simple demonstration exposes the weaknesses of Tesla's camera-centric approach. Nobody doubts that Tesla's Autopilot and FSD have greatly improved over time - but they have also been linked to crashes and investigations. Some experts argue that relying solely on cameras is just not enough for safe and reliable self-driving tech. The human eye, while capable, is not perfect, and cameras can be similarly limited by lighting, weather, and other environmental factors.

Tesla's idea of a camera-only system comes from the belief that it mirrors human driving. Unfortunately, the painted wall test suggests that this approach misses crucial details that other sensor technologies can easily detect. The ability of a painted wall to confuse the system might be unrealistic and exaggerated - sure. But it shows that reliance on cameras alone leaves gaps in the system's ability to safely interpret its surroundings.

Similarly, the Model Y didn't even attempt to brake when it drove through smoke and rain, although it did stop when blinded by the strong light. This is the most obvious sign that Tesla's technology is far from ready for everyday use. We have no doubt the debate over the best sensor technology will continue, and the painted wall experiment is a stark reminder of the challenges involved in creating truly reliable self-driving vehicles.

Source