◆L2 driver assistance accelerates landing, new car companies continue to lead, and traditional car companies are chasing after it.
With the rise of autonomous driving, the driving assistance market, as the cornerstone of autonomous driving, is gradually heating up. With the entry of Tesla and the success of Autopilot driver assistance system, more car companies began to pay attention to the added value brought by driver assistance technology.
It can not only effectively improve the safety of driving, but also shape the technological side of car companies in some aspects. For the current consumer market, scientific and technological ability is an important label to measure the strength of a company, and it can also invisibly enhance the brand content of a car company.
1) Don't buy Tesla for used cars?
From a technical point of view, the driver assistance function does have a certain threshold from hardware to software. Let us take Tesla as an example. Tesla Autopilot driver assistance system first used monocular camera as the main sensor of the system, and its platform was Mobileye? EyeQ3, with Tegra from NVIDIA? 3. In addition, a millimeter wave radar with a detection distance of 160m is installed in the front of the car.
Then there was an accident that collided with a big truck. Since then, it has parted ways with Mobileye, a top vision solution provider, and turned to NVIDIA. The second-generation upgraded Tesla began to use the front-facing three-eye camera we see now, and the autopilot platform was upgraded to NVIDIA's DRIVE? PX2, upgrading the autopilot platform with multiple cameras and higher computing power is the biggest change of the second generation autopilot platform.
And drive NVIDIA? PX2 is upgraded again, which is considered as a half-generation upgrade. At that time, the number of processing chips was increased (from one to two) to improve the computing power, because the camera needed to perceive color and brightness. The most breakthrough is the release of the third generation product HW3.
This is 20 19, and Tesla officially launched its own FSD chip. The new platform is equipped with two self-developed FSD chips, and the FPS has made great progress-20 times (2300FPS) higher than the previous generation, and the computing power has increased by 1 1 times, reaching 144TOPS. But the cost of HW3 is reduced to 80% of that of HW2.5.
Do you know what this means? If you buy a second-hand Tesla a few years ago, even if it is equipped with Autopilot system, it is far from the current product.
Similar to our computer, the hardware has been determined before buying, and it will become more and more stuck with the upgrade of game software in the later period. Only better hardware can meet a higher degree of "picture fluency". Those so-called algorithm upgrades may improve the experience to a certain extent, but they will always touch the ceiling of these hardware.
What is the current Tesla model? This automatic driving platform is adopted. In the previous driver assistance test and active emergency braking test, the new Tesla model? All the results are excellent, and the brand-new autopilot platform has played a very important role.
It took Tesla about six years to upgrade the driver assistance system, which did not include the previous research and development, but the final result was remarkable. Tesla's current driver assistance system is among the best in the whole market.
From relying on external suppliers to independently developing chips and algorithms, Tesla is forming a closed-loop ecology similar to Apple. Its autonomous driving development has little impact on upstream suppliers, but it poses a great threat to other car companies.
2) The new forces in China continue to increase the number of "autonomous driving" (driving assistance).
Tesla, as a foreign new force car-making company, has a great influence on China, and domestic new force car-making companies also focus on "autonomous driving" (assisted driving at this stage).
At this auto show, Tucki, who is playing at home, released his next-generation autonomous driving architecture, while Xpeng Motors said that it would upgrade the driver assistance software and hardware algorithm on the production model of 202 1.
Yu Ding, regional marketing director of Xpeng Automobile Company.
However, the Audi A8 equipped with lidar is obviously more dazzling than practical. The detection range of this low-cost lidar is only about 80-100 m. Because of the existence of long-distance millimeter-wave radar in this car, lidar is more like accurate judgment of objects, similar to the multiple redundancy of the system.
At present, Tucki's XPILOT autopilot system has been upgraded to version 3.0, equipped with 13 camera, 12 ultrasonic radar, 5 millimeter-wave radars, dual-frequency high-precision GPS, and supports high-precision maps and V2X technology. Is the computing platform NVIDIA DRIVE? Xavier, the computing power (30TOPS) of this platform is higher than the Tesla Autopilot described above. 2.5 computing power (24TOPS) level, but with the model? The computing power of 144TOPS of 2FSD chip on 3 is still far from each other.
However, this system can fully meet the L2 driver assistance function at present, and can complete the high-speed autonomous navigation driving function after completing the OTA upgrade early next year.
In addition, Wei Lai and Li, two other car companies created by new domestic forces, are also active in the field of intelligent driving. Weilai's driving assistance function in production vehicles is similar to that in Tucki, but the driving assistance platform is Mobileye? The data processing unit system is based on EyeQ4 chip, and the cameras are all three-eye cameras like Tucki. At the same time, five millimeter-wave radars, four panoramic cameras and 12 ultrasonic radar are arranged around the car body. Weilai's hardware can support L3-level automatic driving, which is more than enough for the current L2-level driving assistance function.
Note: TOPS is a performance index to evaluate the computing power of a processor. 1TOPS means that the processor can perform one trillion operations per second.
At present, there is a gap between Li and He Weilai in hardware. First of all, the types of sensors are obviously less than those of Xiao Peng and Weilai, which leads to the fact that although they can perform the same function, they need manual participation and cannot complete full-automatic operation, such as automatic lane change function. Li Yi needs the driver to judge the rear environment. In addition, because there is no angle radar, the ability to judge vehicles from oblique obstacles should not be weakened. Therefore, the current ideal mainly depends on the iteration of the algorithm to upgrade its intelligent driving function.
3) Traditional automobile brands are close behind.
At this Guangzhou Auto Show, BMW finally opened the automatic lane change assistance function to China. Coincidentally, before the auto show, Mercedes-Benz also demonstrated the new E-class automatic lane change assistance function at the E-class test drive meeting. It can be said that at present, the intelligent driving function of foreign high-end brands in China has covered most L2-level auxiliary driving functions, and some high-end models have been able to support more advanced automatic driving functions in hardware.
BMW iX3 is equipped with BMW's automatic driver assistance system Pro. This system is the latest driver assistance system promoted by BMW in China. This system has been completely upgraded in hardware, including 12 ultrasonic sensor, which is mainly used for parking assistance system. 1 Forward-looking three-eye camera, which is mainly used to shoot images for image recognition; In addition, there are millimeter-wave radars, which are mainly used to detect vehicles and obstacles in front of them.
"BMW's automatic lane change assistance needs to be started above 70km/h"
With regard to this brand-new system, BMW's technical team started the redevelopment and verification of road conditions in China two years ago. BMW's technicians admit that the system is specially optimized for the unique traffic signs, high-speed fishbone markings and obstacles on the road in Gasser and China, and the road conditions in China have also improved BMW's own technical level.
Mercedes-Benz has also demonstrated similar functions in the new E-Class, and the Mercedes-Benz technical team also said that the road conditions in China have been partially optimized and upgraded.
Eternal tower? ADiGO from LX? 3.0 The system is equipped with 12 ultrasonic radar, 4 short-range lateral millimeter-wave radars,/kloc-0 forward medium-range millimeter-wave radar and 5 cameras (of which 1 is mainly used for forward object recognition, and the other 4 are mainly used for realizing 360 panoramic image function). These sensors can be used to detect objects around the vehicle.
Other China brands, such as Geely, showed quite satisfactory driving assistance functions at this auto show. Most models are equipped with standard monocular camera and millimeter-wave radar, which can meet the common L2 driver assistance function in hardware.
However, according to the performance of UNI-T, UNI-K, as a higher level, will never lower the requirements for driver assistance functions, so this car will also cover most L2-level driver assistance functions, and at the same time, a high-precision map may be introduced later, so it should be possible to complete lane change assistance in expressway. Of course, this is all speculation at present.
Summary: Below, let's summarize some situations of auto companies in various countries in this auto show:
Traditional brands in China: In terms of traditional car companies, all brands are promoting L2 driver assistance function regardless of vehicle price. The types and quantity of sensors are obviously increased for high-end models, while the number of sensors for low-end models is less. The function is mainly based on the commonly used driving assistance (adaptive cruise and lane control functions). It can be seen that China brand first solved the problem that all models have driver assistance. From a technical point of view, except for some high-end models, most models have a certain gap with the new power car manufacturers in hardware composition and intelligent driving platform.
In terms of domestic new car manufacturers, "driver assistance function" is still one of the core competitiveness of these brands, so car companies are more willing to invest deeply. In terms of hardware, Tucki and Weilai are redundant and belong to the first group, but their concepts are slightly backward. There is still a big gap between the computing platform and Tesla, but whether the driver assistance function needs L4 computing power at this stage is still a question mark. Generally speaking, the three new domestic forces are at a high level from computing power to hardware intelligence, but there is still a certain gap with Tesla.
In terms of foreign brands, Tesla is leading the way in terms of technical route and functional coverage. Although traditional high-end brands are conservative in intelligent driving, it can be seen that they all have strong technical reserves. At the same time, foreign car companies are constantly absorbing China road information to feed back their driving assistance functions. In addition, BMW, Mercedes-Benz etc. Hardware configuration supporting more advanced autonomous driving has been completed, and more comprehensive driving assistance functions may be introduced later.
◆ The regeneration and change of the driving assistance technology route of mass production vehicles is the biggest pusher behind the cost or success.
At this Guangzhou Auto Show, in addition to the driver assistance function becoming more popular, we found that the technical route of driver assistance has changed again.
1) Pure monocular camera may be the first choice for low-cost vehicle driving assistance.
As we know, the current mainstream driver assistance hardware technical solutions are visual solutions and visual+radar solutions. Pure vision only uses visual sensors, while mixed vision uses both vision and radar.
In the past, it was known in the industry that the pure visual route would gradually decrease, and the route with millimeter-wave radar and monocular camera would become the mainstream (please refer to: Who can see clearly at night? Automatic braking test of three cars at night). But from the observation of this auto show, it is obvious that the pure visual route has not disappeared, but has become popular with the breakthrough of monocular visual technology.
Although it may be affected in bad weather conditions, it is enough to achieve the purpose of car companies in terms of function realization, and the cost of pure monocular camera route is lower, including suppliers such as Bosch and Chinese mainland have completed related products, which is not a problem in the supplier link. It can be predicted that in the future, low-end models will be equipped with pure monocular cameras to realize driving assistance functions.
2) High-end models are still developing multi-sensor fusion, and lidar is beginning to make progress.
If low-cost models will realize the popularization of driving assistance function through monocular cameras, then high-end models will still consider the integration of multi-sensors. In terms of vision, it began to evolve from a single camera to multiple cameras, such as Tesla and BMW mentioned above. They all evolved from single camera to three cameras to enhance visual perception. Manufacturers like Subaru, which started with binocular cameras, began to integrate millimeter-wave radar into the system to improve the stability of the system in bad weather.
In the past, lidar was mostly used in L4-class self-driving test cars, but at this year's auto show, we also began to see top luxury cars equipped with lidar, such as Maibakh S-class. Xpeng Motors also announced that it will be equipped with lidar at the auto show. In fact, some time ago, both Volvo and the Great Wall had the news that lidar got on the bus. With the advantages of high precision and high detection distance, lidar is gradually accepted by car companies. With the rise of Chinese lidar suppliers, the cost of lidar is also decreasing, so it can be said that autonomous driving is accelerating. ?
Editor's comment: Looking back at this Guangzhou Auto Show, we find that car companies pay more and more attention to the concept of driver assistance. Although all the participating models are high-end or top-of-the-line models, from the price point of view, more than 6,543,800 top-of-the-line models have achieved full coverage in driving assistance, and the promotion speed of L2 driving assistance function is faster than we expected.
The monocular camera with pure vision is likely to occupy the low-end driver assistance market in the future due to the upgrading of technology; In the middle and high end, the technology of radar+camera fusion will continue to develop, and the intelligent driving platform will still be dominated by foreign suppliers (such as Mobileye). In addition, lidar is also eager to try, but due to cost constraints, I don't think the time for mass production and loading will come soon. (Photo: Shu Ning Yang Peng? Text/car home? Leng xiaoyang)