access:
The police said that the accident occurred at about 3:30 in the morning when Tesla ran into a truck at an intersection and was driving in an unknown direction. The two Tesla injured were immediately taken to a nearby hospital in critical condition.
The driver of the semi-trailer was not injured.
Can’t see the white Tesla
This is not the first time Tesla has collided with a white truck.
In 2016, a Tesla Model S in Florida, USA, collided with a turning white semi-trailer in the Autolipot state and got under the truck container. The Tesla driver was unfortunately killed.
I thought this was the first auto-driving fatal accident in Tesla’s history, but later it was discovered that a Tesla Model S crashed into a road sweeper in China. The fatal accident happened earlier.
In 2019, still in Florida, a Tesla Model 3 rammed into a white trailer truck that was slowly crossing the road at a speed of 110 km/h. It is understood that the Model 3 was also in the Autolipot on state at that time, but neither the driver nor the Autolipot made evasive actions, so there was a tragedy in which the vehicle was “cut its head” and the driver was killed on the spot.
In June 2020, a Tesla Model 3 driving on a Taiwan highway ran into a white truck that had rolled over on the road. It is understood that the car was in the Autopilot automatic driving system at the time of the incident. Because the white truck is loaded with soft material goods, there is a large buffer space in the event of a collision, so no casualties are caused.
In Tesla’s eyes: the white car box and the sky are the same color
The investigation results of these accidents showed that because Tesla’s autopilot system mistakenly identified the white cargo compartment of the truck as the sky, it caused the truck to crash directly without slowing down at all.
Why is the semi-trailer truck container still white? Don’t you know that you are “invisible” to Tesla.
Some netizens have previously used the following picture for visual recognition experiments.
Import the white truck picture into Photoshop and use the quick selection tool to try to check the outline of the white truck. The result is this:
A large swath of blue sky and white clouds was also marked in the check box. In Photoshop, the white cargo box is the same as the sky.
The results of Tesla’s autonomous driving visual recognition system may also be the same.
Looking at the large white cargo box in the picture of the Detroit accident, it seems that this accident could not escape this reason.
Tesla’s autopilot system mainly relies on visual recognition.
However, the system’s visual recognition decision-making has made fatal mistakes time and time again.
There have been accidents of the same reason many times. I don’t know how Tesla will respond this time?
Prior to this, Tesla has always emphasized that car owners have not kept their attention, and the official investigation concluded that Tesla is not responsible.
Tesla always emphasizes “Autopilot assisted autonomous driving” and “FSD fully autonomous driving” in its publicity. The illusion given to users is that Tesla can achieve autonomous driving. There is only a humble disclaimer on Tesla’s official website: “The currently available (Autopilot) function requires the driver to actively monitor, and the vehicle has not yet achieved fully autonomous driving.”
Recently, in the face of repeated inquiries from the US government, Tesla finally admitted that its Autopilot and FSD functions are only driving assistance functions. Eric C. Williams, Tesla’s vice president of legal affairs, admitted in an email to the government:
“The Autopilot and FSD functions are not autonomous driving systems. Neither alone nor as a whole, neither has the function of automatic driving, nor does it allow our vehicles to achieve automatic driving.”
In short, it is not Tesla’s fault, but the fault of the people on earth.
Tesla is worthy of being the “non-stick pan” in the pan.
Semi-trailer trucks also need to be changed
Observing the close contact between Tesla and the semi-trailer these few times, Tesla rushed directly to the bottom of the truck without slowing down.
In this regard, many netizens have expressed that semi-trailer trucks should be changed in their settings-to install guardrails under the trucks.
And this regulation is not aimed at Tesla. For example, Germany has this regulation before Tesla entered Germany.
After all, accidents like Tesla are not the only ones, and there are many such accidents in other traditional cars.
But careful netizens also discovered that in the Tesla accident in Detroit, the semi-trailer truck was equipped with a protective plate.
But it seems that the protective plate is too soft, and the installation length is not long enough to cover the entire car box, still let the Tesla driving at high speed “take a loophole.”
Regardless of whether Tesla responds or makes improvements in the future, this suggestion is necessary.
The carriage is painted with the words “STOP”? In this case, Tesla’s visual recognition can at least see something, right?
The future of autonomous driving, is it lidar or visual recognition?
The change of semi-trailer trucks is only to reduce the damage as much as possible, and the main problem lies in Tesla’s autopilot system.
In Musk’s eyes, lidar is a backward thing. He once said in April 2019: “Lidar and high-precision maps are both wrong solutions for autonomous driving, which will only slow down the commercialization of autonomous driving. Speed. If you are only driving in a restricted area, it cannot be called autonomous driving.”
In contrast, currently Toyota, Ford, and General Motors are all betting on lidar, which is clear information, which is two completely different routes from visual predictions.
Lidar scans a wide range through active detection, and because the laser beams are more concentrated, the accuracy is higher than the combination of a camera and a millimeter wave radar.
The main disadvantages of lidar are that it is big, heavy, and ugly, while Tesla only needs to install a small camera and a few millimeter wave radars.
The current mainstream view is that lidar measurement is the most accurate, while Tesla’s computer vision recognition is still slightly less accurate.
Before autonomous driving is popular, let’s learn to drive!
Reference materials:
https://www.huxiu.com/article/360888.html
https://www.radio.com/wwjnewsradio/news/local/tesla-crushed-underneath-semi-in-southwest-detroit-crash
https://zx.sina.cn/abc/middle1.d.html?type=auto&link=http%3A%2F%2Fauto.sina.cn%2Fzz%2Fhy%2F2021-03-12%2Fdetail-ikkntiak9384657.d.html%3Fsinawapsharesource%3Dnewsapp&wm=3200_0002
.