For Some Reason, People Are Trying to Crash Teslas Into Walls. It Makes More Sense Than It Sounds

A Tesla Model Y, a wall simulating a road, a YouTuber with 65.8 million followers, and his reputation. Tesla’s reputation, its safety systems, and semi-autonomous driving. Another YouTuber in a Tesla Cybertruck and another wall with another painted road. People are throwing their Teslas against a wall. What’s going on?

The game of Wile E. Coyote and the Road Runner. “I’m in my Tesla on Autopilot going 40 mph toward a fake Wile E. Coyote and the Road Runner painted wall,” that’s how Mark Rober, a well-known NASA engineer with a YouTube channel boasting 65.8 million subscribers, starts a video.

Rober drives his car into a fake wall, simulating a road continuation, and perfectly mimics reality: a set. The goal is to test whether his Tesla Model Y, equipped exclusively with cameras, performs better or worse than a car with a LiDAR system in terms of autonomous driving.

The background. In the video, Rober explains the differences between a car relying solely on semi-autonomous cameras and one using LiDAR. In 2021, Tesla announced it would base its driver assistance system on cameras. In 2022, the company said it would stop installing radars and sensors.

The company claims its software can analyze everything using cameras alone with reliability equal to that of radars and sensors. Of course, this also saves the cost of installing these components, which increases the final cost of the vehicle.

To test this, Rober uses a Tesla Model Y, which relies only on cameras in its Autopilot system, and a Lexus RX with a LiDAR sensor from the company Luminar. The cars are driven with their semi-autonomous features enabled (Autopilot for the Tesla and Adaptive Cruise Control for the Lexus) to see what happens in different scenarios, including children appearing in the car's path or very heavy rain.

The wall. During the tests, the LiDAR-equipped vehicle slowed down. It saved a child from being hit if it was already on the road, appeared behind an obstacle, was in fog or heavy rain, or was blinded by intense light. And yes, even in front of a fake wall simulating a road.

The Tesla Model Y couldn’t say the same thing, as the cameras failed the fog, rain, and famous wall tests. In the latter case, the YouTuber tricked the car into thinking the road continued and didn’t stop.

The controversy. Rober’s video has sparked debate on social media platforms for several reasons. The first and most obvious question: Does this test make sense? How can a car hit a wall that completely covers a road and pretends to go on?

Aside from that, what’s certain is that Rober activates the Autopilot a few seconds before reaching the wall. In the video, you can see the system deactivate when it’s about to crash. Users on social media accused Rober of turning off the system. He defended himself by publishing a complete video showing how the system was automatically deactivated moments before the crash. For that reason, users have criticized Tesla, accusing the company of shirking responsibility to authorities.

More from Xataka On

Researchers Disassembled Batteries From Tesla and BYD to Find Out Which One Performed Better and Cost Less. The Answer Might Surprise You

Rober was also criticized for using a vehicle with Autopilot and not Full Self Driving (FSD), Tesla’s most advanced autonomous driving system, while the Lexus was equipped with LiDAR, one of the most expensive technologies available today.

Cameras vs. LiDAR. The experiment is intriguing, even if it’s not conclusive. You might expect a camera-based system to be less effective than one with LiDAR. However, the combination of camera and radar has proven to be the most effective option so far, both day and night.

Cameras analyze what’s in front of them with images, but LiDAR constantly sends pulses of light around it. This light bounces off objects and returns to the LiDAR. The time that elapses between sending the light pulse and receiving it back can create a 3D map that, in the case of the wall, checks if there’s an obstacle in the way, even though it may fool the camera.

Autopilot and FSD. Offered as an add-on option on a subscription basis, FSD allows the driver to take their hands off the wheel (even in urban environments) as long as they pay attention to what’s happening in front of them.

For example, Autopilot requires the driver to acknowledge lane changes, while FSD navigates between streets on its own without the need for driver acknowledgment. In the U.S., cars can travel 0.40 miles without a driver. So, you could ask a parked Tesla to pick you up from a distance.

Some people drive their cars into walls. Rober’s experiment has inspired other YouTubers to defend the company. In response, Kyle Paul, whose YouTube channel has 237,000 subscribers and specializes in Tesla videos, posted his own test.

The goal was the same: Would his Tesla Cybertruck break in front of a wall simulating the continuation of the road? He answered that if the car were on Autopilot, it wouldn’t hit the wall—or so it seems, because, again, there’s controversy.

The other controversy. Is it true that the car wouldn’t hit the wall? The problem is that Paul did his test in the late afternoon. Not only is the reproduction of the road on the wall of lower quality, but it’s also easier to attract the cameras’ attention.

In this case, the ambient light will most likely help the cameras recognize that there’s an obstacle in front of them and detect it more clearly. This is especially true when you consider that the ambient light is already poor enough that you have to activate the car’s cameras.

Does all this matter? Yes and no. No, because if the point is to prove that a LiDAR system works better than cameras (as long as no other devices accompany them), users already know that, and it’s been proven. The real challenge for Tesla is to prove that its AI systems and cameras can do the same job for less money.

On the other hand, Tesla wants to put driverless cars on the road in a robotaxi service powered by its most advanced autonomous driving systems. Until recently, authorities were clear that Tesla wasn’t at the level of Mercedes, which can offer level 3 autonomous driving (doesn’t require driver attention in some situations). However, Tesla has already received the go-ahead from California to test its cars on the road, as long as it does so with its own employees.

Image | Charlie Deets (Unsplash)

Related | Tesla Recalls Nearly Every Cybertruck Ever Sold in the U.S. It Has a Good Reason: Some Parts Are Falling Off