The delivery time is concentrated from the end of 20021to the beginning of 2022, so there is basically no problem with the title "lidar gets on the train from 20021".
Then I remembered an amazing sentence that Musk said a few years ago:
"Lidar is stupid, whoever relies on lidar will be finished. (Lidar is a fool. Anyone who relies on lidar is doomed to failure.
It can be said that it is harmless, but it is extremely insulting.
How to evaluate Tesla's release of fully automatic driving hardware, and said: Only fools use lidar? www.zhihu.com
According to SAE standard, autonomous driving is divided into six levels: L0, L 1, L2, L3, L4 and L5. L3 is the most difficult to cross, because it involves the key question "Is the driver or the car responsible for the accident?":
When can we achieve real autonomous driving? What's the difficulty?
Before L3, which is the most difficult to cross, manufacturers have been circling around L2 function for many years. Without relying on lidar, they have made L2 classic functions such as ACC and LCC more mature, and some enterprises have even put high-speed driving assistance functions into practice (Tesla NOA, Tucki NGP, Weilai NOP).
So, now that the perceptual killer lidar is provided, does it mean that L3 and L4 will be realized soon? Or, if there were no L3 and L4, would lidar not be needed at all?
The answer to these two questions is no!
Just as Model 3 is equipped with HW3.0 hardware, but FSD Beta is still far from home, these models equipped with lidar can only be said to be "hardware supports L4-level automatic driving". As for when users can use L3 and L4 to drive automatically, it depends on the speed of algorithm development, the progress of data collection, the regulations and whether you can pay for the software.
So, is lidar useless for the time being? No, because lidar is not only useful for L3 and L4, but also can significantly improve the functional experience of L2 assisted driving!
It stands to reason that there are security risks in both cases. Why is my reaction completely different?
The reason is that I know a little about the capability boundary of the sensing sensor: I have confidence in the monitoring function of the BSD blind area composed of millimeter-wave radar and camera, but I am not so confident about whether they can timely and accurately sense congestion in various scenes (including night and strong light).
To be clear: even if the car can successfully cope with the traffic jam scene of 100, it can't let me relax my vigilance-unless I can add the perception ability of lidar.
Why is lidar more reassuring? This should start with its principle:
In the past, a laser was straight, which was equivalent to digital scanning. Theoretically, you can clearly know what the surrounding environment is like by scanning all the points around you.
Unlike lidar, the camera collects pixel information, which is similar to what the human eye sees.
Different from people, the human eye is equipped with a super intelligent processor (brain), which can easily identify lanes, vehicles, pedestrians and so on in the environment. For vehicles, pixel information is just a bunch of meaningless numbers, which must go through the following complex processes such as abstraction and reconstruction, and must rely on super intelligence to achieve human recognition effect.
In other words, if there is no lidar, it takes 10 times more effort to fill the boundary of perception ability through intelligent algorithms.
For another example, when we human drivers are driving, we occasionally look at presbyopia. For example, near objects are recognized as far away-just as the brain produces optical illusions when processing pixel information.
But if each pixel is marked with distance information (equivalent to being equipped with a laser radar), it is impossible to produce this optical illusion [1]!
In addition to the above-mentioned scenes, lidar will be very helpful to improve L2 function experience in the scenes of bright light change, corner cruising, night driving and narrow traffic.
Musk called lidar a fool, and I'm afraid it won't be easy to change for a while. I don't think it will work.
So, why doesn't he use lidar? There are two different ways of understanding:
At first, a laser radar cost hundreds of thousands of RMB, but I couldn't afford it if I wanted to use it-since I couldn't afford it, I called it an idiot by the way. He probably didn't expect that in this magical land of China, Huawei and DJI (a subsidiary of Lanwo Technology) could reduce the cost of vehicle-mounted lidar to 1000 yuan so quickly with their R&D and manufacturing capabilities.
What is the concept of thousand yuan level? That means, even if one day in the future, you can really achieve L3 and L4 level automatic driving only by relying on the camera, then it is cost-effective to add lidar as redundancy to further ensure safety!
Moreover, there is still room for further reduction in this cost.
As consumers, life is the most important thing. Even if one day the automatic driving of L3 and L4 can be realized without lidar, I still hope to add lidar as redundancy to ensure foolproof.