2023 can be said to be the first spring for smart driving, when consumers began to vote with their wallets, and the penetration rate of smart driving accelerated faster than expected.
According to the research report on smart driving penetration released by Tianfeng Securities, in February 2024, the penetration rate of standard L2 and above intelligent driving was 44.
57%, which was nearly half, an increase of 10.
94% compared with the same period last year, an astonishing growth rate.
From the point of view of the sub-price range, nearly 80% of the new energy passenger cars on the market with more than 200000 yuan are equipped with L2 and above smart driving functions.
Sure enough, as soon as the configuration came up, the price was knocked down, which can be seen in the volume of this year’s car market.
Consumers’ enthusiasm for smart driving is not without warning.
When Xiaopeng G6 was launched in June last year, high-end configurations once accounted for more than 50% of sales.
in July, the new M7 was closely followed by the launch of the market, and consumers who chose Huawei chose Huawei’s level of high-end smart cars.
Official data show that in the two and a half months since the launch of the new M7, the total number of M7s has exceeded 100000, of which the smart driving version accounts for 60%, the optional rate of urban smart driving reaches 75%, and the recommended value of NPS for smart driving is more than 87%.
That is to say, in the era of smart electricity, the high-level intelligent driving function has become a basic configuration here.
Even among the professionals in the car circle, there may have been a misconception that the advanced intelligent driving function, especially represented by the urban NOA, is the exclusive identity of the new power of car building.
In fact, since last year, almost every Chinese car company has been frantically promoting the high-level smart driving function city NOA to get on the car.
BYD announced at the “2024 BYD Dream Day”: “in 2024, BYD will launch more than 10 high-order intelligent driving models equipped with lidar, and high-priced smart driving will become the standard for more than 300000 of its models.
” Wang Chuanfu is really true to his word, from the momentum N7, looking up to the U7, to the upcoming 2024 Han, a series of new BYD models have provided high-level smart driving functions.
Although the landing speed is not as fast as Huawei Xiaoli, it is said that the first batch of smart cars carrying urban NOA will also be delivered by the end of March, and the first batch may be opened to a number of major cities, including Shenzhen.
And SAIC’s Zhiji, Geely’s polar krypton, Great Wall’s WEY, these traditional car manufacturers, all chose to fight “urban NOA” without exception.
Just last month, polar krypton officially announced that the new extreme krypton 001 lidar is standard.
Users of the new extreme krypton 001 will open the ZAD fully intelligent driving system for 0 yuan before March 31, including urban highway NZP and urban NZP.
All intelligent driving functions are free for life.
“all of a sudden, it’s like a spring breeze blowing overnight, like a thousand trees and ten thousand pears in full bloom.
” No wonder Chen Yudong, the former president of Bosch China, said in a live broadcast not long ago: “within five years, the urban NOA will infiltrate into the 100,000-yuan level of cars.
” What about foreign car manufacturers? Some said they didn’t understand it, while others panicked.
As the Beijing Auto Show is about to begin, Toyota’s invitation to the automotive media reads: “this year, Toyota will focus on intelligence and bring you important information and new proposals.
” What is the specific plan? It is reported that Toyota’s global smart driving program will adopt the tripartite joint scheme of “Toyota + Huawei + Momenta”.
Some media reported that suppliers of Toyota’s overseas smart driving orders include Desi Siwei, Qualcomm and Momenta.
The three companies each assume different roles and jointly complete the development of Toyota’s intelligent driving system for overseas projects.
While the big German factories do not seem to be particularly fond of the “half-baked” urban NOA, people have defined high-level intelligent driving above L3.
At the end of last year, China began to issue L3 test licenses, and the first one to win the test license was BMW, which has always kept a low profile, and finally found an opportunity to prove its technological strength to the outside world.
Like BMW, the layout of the German Big three in the field of smart driving has been held high, anchoring the L3 to target the high-level intelligence of the car company, hoping to deliver a smart driving function that can be called “easy to use”.
Even if this feature can only be enabled in low-speed scenes on highways.
In the field of high-level smart driving, ABB all chose the lidar scheme without exception, and the person in charge of the intelligent driving development of the German car factory even believes that in the future, because the definition of intelligent driving in different consumer markets is too different, the product form of lidar will have absolute differences in parameters and shape.
Just as the vehicle camera also has differences in forward-looking, fisheye and circumferential vision, the lidar will certainly have a category subdivision similar to that of the camera in the future.
in order for the forward-looking lidar to truly achieve over-the-horizon, the ideal detection distance needs to be more than 300m.
The round-looking lidar needs to cover a distance of about 50m, and its important function is to solve the fast avoidance in low-speed crowded scenes.
In Europe, the advanced intelligent driving function starts from the L3, with the high-speed as the core scene.
At present, the L3 function can only be enabled in low-speed scenarios, precisely because there is no ultra-long-range lidar product suitable for European high-speed L3, and there is no speed limit for more than 60% of highways in Germany.
With this in mind, I believe people will no longer make fun of the cautious L3 function of the Germans.
Looking back at the Chinese market, how much of the high-level intelligent driving function represented by urban NOA has been realized as a hot spot of various car companies competing for PR? Yu Dazui said that by the end of 2023, “all over the country can be opened, all can be used” urban NOA has been realized or not? What’s the difficulty? Let’s start with the conclusion.
Yu Kai, founder and CEO of Horizon, said publicly at this year’s 100-person forum that high-speed NOA has just arrived “easy to use”, less than consumers “like to use”, and urban NOA is not “available”.
This can be said to be the most “sincere” summary of professionals.
Let’s first take a look at the test results of the horizontal review.
just last month, we released the latest horizontal evaluation report of the urban NOA, selecting the models that claim to have (or will soon) promote the NOA function in the urban area, namely, the new M7, Xiaopeng G6, Lulai ES8 and ideal L9.
Of the 108km test route, only one model can fully enable the urban NOA function, that is, the new M7.
The performance of the rest of the models is uneven.
In the second test, we chose a 180km road in the urban area of Shanghai, and various NOA companies also took over many times.
M7 also had 4 downgrades and 26 takeovers in the test, with an average takeover rate of 7km/.
Xiaopeng G6, which has been promoting the urban NOA, reached 55% of the functional activation rate in this test in Shanghai.
Which demons are they?The scene makes the urban NOA downgrade frequently? To sum up briefly, the scenarios that lead to the withdrawal or downgrade of NOA in urban areas are roughly divided into three categories: first, it is still very difficult for vehicles to identify and respond to dynamic targets, especially those that are close and moving fast, such as pedestrians and electric donkeys.
The conservative smart driving strategy will remind the driver to take over when there is an obstacle, but at this time the time left for the driver is very limited, leaving the pot to the driver at the last minute.
There are even strategies that remind you to take over without slowing down, which is dangerous.
In this kind of scene, if perception can be identified as soon as possible, it will not challenge the limit of human response.
The identification of the dynamic target-electric donkey is not fast enough, and the dynamic target-pedestrian obstacle avoidance fails.
The above two scenes are the car’s intelligent driving function reminding to take over, while still moving forward, the danger can be imagined.
The second category is that there are problems in the identification of static targets, such as fences on highways, cones and triangular warning signs that can be seen everywhere in cities.
At this time, the smart driving function is often “foolish” or “extremely reckless”: either waiting to leave or directly bumping into it.
For example, in the video below, when you get on the ramp, you rub it directly on the guardrail of the highway.
If you follow the car in the rear, you will probably have a rear-end collision.
Unable to identify the high-speed fence, direct high-speed rubbing, even including the APA autonomous parking function that people usually think is safe, but stop at the worst, and will never hit it.
In fact, there are many accidents.
The sensors that can be used in APA scenarios are extremely limited, and basically use fisheye cameras to challenge all kinds of APA parking scenes, which is a lot of pressure for algorithm students.
The camera itself is a two-dimensional sensor, and there is no depth information, coupled with fisheye camera for object recognition also needs distortion correction, the superposition of various algorithms, often more than ten centimeters of error, has done its best.
However, for the narrow parking spaces and oversized cars in many cities, the error of more than ten centimeters is the main culprit for all kinds of parking crashes.
The garage APA parking front crashed, the garage APA front crashed, and then there is the common problem of light map NOA, that is, the ability to identify road edges and roadblocks in the city is limited, and the function may not be degraded, but the experience is complicated.
It can be called an algorithm engineer’s nightmare, and the intelligent driving tester has to scold his mother.
All kinds of “roadside”, on the one hand, the car companies are desperately trying to popularize the urban NOA to every car, on the other hand, the actual functional performance of the urban NOA is not satisfactory.
At present, “BEV+Transformer” has just arrived in the industry from a prospective research paper, and when it comes up again, it feels a bit out of date.
Intelligent driving engineers have just finished learning the latest research results of machine vision, and they are constantly rushing to “end-to-end” and “big model”.
After all, there is Tesla in front, and “end-to-end” claims to have got on the bus and Xiaomi behind the OTA.
, claiming to be entering the first camp of smart driving by the end of this year.
The wolves are surrounded by adversaries.
If you take a closer look at the first camp and the second camp that followed, no one can abandon the lidar completely.
Whether it is BEV or occupying the network, domestic smart driving schemes have been integrated with lidar.
Lidar is not only a redundant backup in the intelligent driving function, but also can not be replaced by the three-dimensional perception of its output, especially the depth (distance) information.
Especially today, the light map scheme is popular, lidar with GNSS+IMU to achieve vehicle self-positioning, will be another main function.
At present, people pay more attention to whether the cost of lidar can be reduced a little bit.
Wang Chuanfu also claims to develop his own lidar, and the cost hopes to be controlled at around 900 yuan.
of course, if you look carefully at what he said, the complete message he conveys is: “not the kind of high resolution, but the kind of lidar with enough performance.
We think it can be controlled at around 900 yuan.
” While waiting for the industry to come up with cheaper components, redefining the parameters of the sensor from the actual needs and finding a sufficient and easy-to-use lidar is likely to be another path to lidar success.
And this coincides with the German car company’s definition of sensor that I mentioned earlier.
Urban NOA requires vehicles to output close-up perceived obstacles as early as possible.
as long as they block the way, the vehicle will slow down first, not to remind the take-over of the speed output at the same time immediately before the collision.
This is the perceptual logic of lidar.
Whenever an obstacle is detected, no matter what object it is, it will feedback “unknown big” or “unknown small” to the system as soon as it is detected.
To put it simply: don’t ask what it is, slow down first.
Compared with camera perception, no matter what large model you use, you have to first identify what it is, and then give feedback to the system.
For urban NOA, lateral lidar is also an important option, which can be accurately sensed by real-time output within 50 meters, the large vertical view angle, the crossing of close pedestrians and the serpentine alignment of takeout boys can be detected in time to avoid dark under the lights.
The lidar is installed on the roof of the car, and there is no point cloud within 5 meters of the body, so there is no lidar perception, so it is in the perceptual blind area.
In particular, there are all kinds of flower beds, green belts, isolation belts and stone piers in the city.
China is so large that the obstacles are endless.
With lateral lidar, the cost of these training can be saved, and it is one step closer to the urban NOA in the light map.
At present, some car companies have begun to get on the bus with the layout of lateral lidar, which is different from the “Don’t talk with less than four” two years ago.
The return of lateral lidar is precisely the rational choice of many car companies after they are calm.
Now it’s time for the Beijing Auto Show and the critical moment for the annual drama of the new car launch.
which new models will be added to the urban NOA function this year? Will everyone continue to roll “Kaesong” at the same speed? Will you pay more attention to the number of takeovers per 100 kilometers of your products? Will it bring us closer to the “available” urban NOA features? Beijing Auto Show, let’s wait and see.
, return to the first electric network home page >.
