Tesla Autopilot was wrong and smashed into the wrong wall into the wrong wall into the wrong wall

Tesla Autopilot drove into a Looney melody in the fake road in the middle of the street. This was the result of a camera against the Lidar test for autonomous driving from Youtuber and the former NASA engineer Mark Rober.

In this video, Rober puts a Tesla model Y on autopilot against a vehicle with a lidar system (Light Detection and Ranging) in various tests.

In addition to other discoveries, the video showed that the Lidar sensors didn't care what was painted on the wall. They only took care that it was a wall. Cameras, on the other hand, were slightly equipped.

Elon Musk believes that only fools use lidar

Elon Musk was an open critic of laser-based recognition, Lidar for self-driving technology and even in 2019 called “crutch” and “fool's care”. Most companies that develop self -driving technologies have used a mixture of sensors such as cameras, lidar, radar and ultrasonic.

The Tesla CEO believes that as soon as you solve autonomy, you will scale faster as competitors, since its vision plus neuronal network system is designed in such a way that you can work like a human driver and therefore adapt to any street. He argues that Lidar implements the core problem of visual detection and conveys false progress and believes that real autonomy requires almost perfect computer vision.

Experts, on the other hand, have pushed back against these claims. Musk has mentioned that Tesla reaches the “Autonomy of Level 5”, which means that “complete automation can carry out all the driving tasks under all conditions without human interventions” that cameras can easily cover the LIDAR sensors.

“Lidar is a fool's job,” said Elon Musk. “Anyone who relies on Lidar is doomed to fail. Taldied to fail! [They are] expensive sensors that are unnecessary. It's like a whole series of expensive attachments. An attachment is bad, now you have a lot of it, it's ridiculous, you will see. “

In addition, the electric car manufacturer removed radar devices from its vehicle line -up and deactivated radar devices already installed in existing vehicles. This strategy has not yet paid off, since Tesla's systems are still on the driver assistant systems of level 2.

The actual test and why Lidar is top

The laser-based lidar system was much better in recognizing objects in heavy fog or rain than a camera system. The Tesla autopilot has managed to move in tests in which the object was statics, moved or was blinded by lights. However, it could not hold against fog or heavy rain and shattered directly through the window of the window.

However, the most striking test comes later when Rober and the team use the wall. The wall in the Wile E. Coyote style with a wrong street was painted to look like the street, the sky and the surrounding area.

It is unlikely that someone will be exposed to this scenario in real life, but this test illustrates the problem with cameras based on the perception of obstacles and not on hard data, in contrast to radar or lidar sensors.

The vehicle equipped with Lidar recognizes the wall and easily holds it, since it only perceives the wall as an object and not as the painting. The Tesla unfortunately breaks through directly.

“For the first time in the history of the world, I can definitely say that Tesla's optical camera system would throw a fake wall through the brakes absolutely without a light tap,” said Rober.

Tesla's autopilots and complete self -driving systems have improved over time, but remain associated with several accidents, some are even fatal.

However, many autonomy experts claim that Tesla's camera-based approach that Musk believes that he is outstanding, since it is similar to the drive of cars of humans, always have its defects, especially if the goal is to make self-driving cars without steering bikes or pedals more securely than humans.