Tesla’s new version of the “fully autonomous driving” system completely abandons the radar. Is it a “disruptor” or a “killer”?

For a long time, Musk has been promoting his FSD (Full Self Driving).

He once stated in 2018 that the "long-awaited" new version of the FSD system will be launched in August. As you all guessed, the new system did not arrive as expected. In 2019, Musk did the same thing again, drew a "big pie" to fans, which was quite a bit of "wolf coming".

▲Elon Musk, picture from: Wall Street Journal

Earlier this month, Musk finally admitted: The road to autonomous driving is much more difficult than he thought. But a week later, Tesla released the Beta 9.0 version of FSD. This update is considered to be an important step for Tesla to fulfill its promise of "fully autonomous driving" of vehicles.

The new version of the fully automated driving system uses Tesla Vision, a vision system that only relies on optical images, and does not require any data from radar sensors. Prior to this, radar sensors were an important part of Tesla's sensor suite.

▲Picture from: Torquenews

Abandon the new version of the radar FSD system, how is the experience

Tesla officially announced that this update is limited to two thousand car owners who participated in Tesla's "Early Experience Program", and they can already experience the new FSD Beta 9.0.

Although the system still does not achieve true "fully automatic" driving, it still has a number of urban road assistance functions, such as changing lanes or turning, etc. Previously, this function was limited to highway use.

Some of the car owners have already released the experience video of this new version of FSD.

▲Picture from: Chuck Cook

In the video posted by @Chuck Cook, his Tesla can drive on city roads with clear lanes. Red stop, green light, and identification of speed limit signs are already basic operations.

▲Picture from: Chuck Cook

When passing some intersections with complex traffic conditions, the vehicle will automatically slow down.

▲Picture from: Chuck Cook

Not only that, even on small roads without lane lines, the FSD system still has no problems. Cook's Tesla can accurately identify and avoid vehicles parked on both sides of the road.

It is worth noting that not only during the day, Cook also tested at night. From the results, the night is not a problem for this system.

▲Picture from: Chuck Cook

On the other hand, Tesla's "Mind of Car" driving visualization view has also been greatly improved, not only can display clear lane lines, but also the surrounding object detection looks very accurate.

Tesla said that the current "Mind of Car" can better support fully automated driving functions. When the driver activates this function, the LCD instrument will accurately display the road condition information around the vehicle

In the new UI, the system uses red lines to mark the road shoulders, and the white and yellow lane lines are also well restored. The previously updated traffic lights, stop lines, and speed limit signs are also retained.

▲Picture from: Tesmanian

In addition, the system will use different models to distinguish different types of vehicles, pedestrians, and animals, and even identify the status of the car lights. The display effect is very intuitive.

▲Picture from: The Driven

As you can see from the picture, the applications on both sides of the dashboard will be temporarily closed, and the driving speed, Autopilot availability, Autopilot setting speed, and detected speed limit information are all moved to the bottom of the dashboard.

▲Picture from: electrek

Of course, this is only limited to Model S and Model X equipped with LCD dashboard. For Model 3 and Model Y without a separate dashboard, the above information will be displayed on the central control screen.

Interestingly, Tesla is using neural networks and machine learning to improve FSD. So in theory, as more and more people use this system, it will become more and more perfect.

▲Picture from: Tesla

There is still a long way to go before "autonomous driving"

Although in Cook's video, this FSD system performed well overall, it may appear to hesitate when turning at intersections for safety reasons .

Especially when Tesla tried to turn left into a six-lane highway, it waited for a very long time. Finally, Cook intervened and manually drove the Tesla onto the highway.

But objectively speaking, this is indeed a difficult operation. Even a human driver needs to "look left and right" and drive in carefully.

▲Picture from: Chuck Cook

Autopilot human-computer interaction researcher Liza Dixon (Liza, Dixon) said: "Tesla has been conducting beta testing on FSD."

The FSD beta version has been updated regularly, but it should be noted that these updates did not make the system more "autonomous."

In other words, even with FSD Beta 9.0, the driver must put his hands on the steering wheel, ready to take over the vehicle at any time. Musk also said in a tweet: "Beta 9 solves most of the known issues, but we don't deny that there may be many unknown issues, so please be extra careful."

▲Picture from: Electrek

The problem is that drivers who have their hands on the steering wheel and are always alert must not only pay attention to road safety, but also be alert to errors in the FSD system.

This sounds more stressful than driving by yourself, especially since Tesla has warned in advance that the system may "do something wrong at its worst."

▲Picture from: Tesla

As mentioned earlier, the current version of FSD Beta 9.0 only relies on the camera to identify the surrounding environment. Tesla did not use newer and better sensors to enhance its sensing system like other automakers, but instead took all the bets. "Betting" on the camera.

Whether this sensor system is stable enough, it will take time to test. But it should be noted that in this process, not only Tesla owners, but also passing vehicles and pedestrians are taking risks.

▲Picture from: TechNews

At the 2021 World Artificial Intelligence Conference on July 8, Su Qing, President and Chief Architect of Huawei's ADS Intelligent Driving Product Line, even used the term "killing" to evaluate Tesla's autonomous driving:

In the past few years, Tesla's accident rate is still quite high, and the types of accidents from the first person that "killed" to the person who recently "killed" are very similar.
In a limited number of years, or even ten years, the machine will definitely encounter some extreme scenarios that it cannot handle.

How to ensure the safe landing of autonomous driving is an issue that major manufacturers need to focus on.

▲Su Qing (third from left), picture from: 2021 World Artificial Intelligence Conference

In this regard, the establishment of "autonomous driving pilots" may be a more responsible method.

Last weekend, Baidu Apollo Robotaxi officially opened to the public in Guangzhou, becoming the largest autonomous driving service in Guangzhou, with 237 sites, and the number is still growing.

In addition to Guangzhou, Baidu Apollo has also opened up autonomous driving travel services in the three cities of Changsha, Cangzhou, and Beijing. By the end of last year, Baidu Apollo had provided travel services for more than 210,000 people.

▲Picture from: Baidu Apollo

Although Tesla's FSD Beta 9 is an important milestone towards fully autonomous driving, we still have some time before the "completely let go" day.

During this period of time, in addition to directly granting users the right to use "autonomous driving", Tesla should also have more in-depth thinking about the term "safety".

#Welcome to follow Aifaner's official WeChat account: Aifaner (WeChat ID: ifanr), more exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo