The limits of Tesla’s Autopilot system, which relies solely on cameras and AI, have been put to the test by Mark Rober, a YouTuber, engineer, and inventor. In a series of six different scenarios, Rober compared the performance of a Tesla Model Y equipped with the company’s proprietary Tesla Vision system to a vehicle using LiDAR technology. The results were revealing, to say the least.
Tesla has developed its own Tesla Vision system to power its Autopilot feature, which uses a combination of cameras and AI to navigate the road, rather than relying on LiDAR or sensors from other companies. Elon Musk has touted the system as being able to learn and adapt to any type of road, just like a human, and has claimed that it will enable the development of fully autonomous driving capabilities faster and more easily than its competitors. However, the current Autopilot system is still only at level 2, meaning it’s an assistive technology, not a fully autonomous one.
Critics have argued that the system has limitations, particularly given Musk’s previous claims that Tesla is working towards level 5 autonomy, where the vehicle can drive itself in all environments. Some have suggested that relying solely on cameras may not be enough, and that LiDAR technology, which can create highly accurate 3D maps, might be a better solution. However, Musk has been adamant that LiDAR is not necessary.
In Rober’s tests, the Tesla was able to stop for stationary and moving dummies, as well as in the presence of interfering light, but struggled in situations with heavy fog or rain. In contrast, the LiDAR-equipped vehicle was able to stop in all scenarios, thanks to its ability to detect objects more accurately than the camera-based Tesla Vision system.
One particularly interesting test involved Rober placing a fake wall with a painted road image in front of the vehicles. The LiDAR-equipped car stopped immediately, recognizing the wall as a solid object, while the Tesla failed to detect the wall and drove right into it, fooled by the fake road image.
These tests highlight the limitations of camera-based systems and suggest that Tesla’s Vision system may need to be improved, particularly when it comes to detecting objects in various scenarios. However, it’s also worth noting that the tests may have been designed to highlight the system’s weaknesses, rather than showcasing its full capabilities.