As we all know, in the automatic driving technology, Huawei and other domestic manufacturers insist on using lidar, while Tesla is a pure visual faction that insists on using cameras and insists on not using lidar. Even in order to indicate that the visual scheme is the direction, even the millimeter-wave radar used before was removed and only the camera was used.
At this point, Tesla's fans are also relishing it, saying that this is because the end of the visual solution is bionic, and the camera can eventually replace the human eye, which is much stronger than lidar, low cost, and excellent solution.
However, as Tesla canceled the radar and only used cameras, everyone found a fact, that is, Tesla's "ghost brake" incident has soared.
According to media reports, in the past 3 months, Tesla's complaints about the "ghost brake" problem have surged to 107. In the first 22 months, there were only 34 cases, and U.S. regulators have now launched an investigation into Tesla's problem.
The surge in these complaints coincided with Tesla's removal of millimeter-wave radar, so there's reason to believe it's because of it.
What is "ghost brake" refers to the fact that when Tesla users use the autopilot function, the vehicle will suddenly and without warning in some seemingly non-braking situation, as if encountering a "ghost", so there is this saying.
As for why the brakes are braked without warning, the reason is that the automatic driving system FSD misjudged, and it feels that there are people or other obstacles in front of the car and must brake, so it automatically brakes, which will put the user in danger of being rear-ended and other accidents.
Why the misjudgment, professionals believe, is because the vision scheme, did not introduce multiple sensors for cross-validation, so it will go wrong.
For example, the simplest example, behind the bus in front, there is a life-size photo printed, and Tesla is behind, and the camera will see this photo and may think that it is a person.
If millimeter-wave radar, or lidar, it will be a camera, radar multiple verification will be concluded, you will know that this is not a person, but only when the camera, it is possible to misjudge.
When multi-sensor cross-verification, there will be another problem, that is, if the results of multiple sensors are different, who will listen to them, which is also one of the problems that plague many car companies.
Of course, whether in the end is the visual program is more cattle, or the radar program is the most cattle, or multi-sensor together to verify the best, overall, whether it is lidar, or visual solution, there is still a long way to go, we do not have to believe too much in automatic driving, after all, it is still L2 level, can only be auxiliary.