Communication group | To enter the "sensor group/skateboard chassis group" please add WeChat ID: xsh041388
Communication group | To join the "Auto Basic Software Group", please add WeChat ID: Faye_chloe
Remarks: Group name + real name, company, position
With the development of the autonomous driving industry, a number of excellent tier1 suppliers have emerged in China, and their products can be found in many vehicles today, including sensors, domain controllers, algorithm solutions, and so on. In 2020, for the first time, local suppliers appeared in the top ten list of passenger car ADAS market share. However, just a few years ago, the passenger car ADAS pre-installation market was almost all occupied by Global Tier1.
A few years before the establishment of local suppliers, Global Tier1 has been deeply involved in this field for many years, and has a firm advantage in both technology and customer relations. So, how do these local autonomous driving Tier 1 suppliers break through?
Standing at this point in time today, we can clearly observe that many suppliers, such as Neusoft Reach and Jimu Intelligent, etc., started from the commercial vehicle market, and after obtaining mass production projects and cash flow, they gradually moved to passenger vehicles. market expansion. So, why does such a situation occur?
Next, this article will analyze the following aspects:
The development path of local suppliers from the commercial vehicle market to the passenger vehicle market
Similarities and differences in technical aspects of assisted driving between commercial vehicles and passenger vehicles
How do suppliers apply their previously accumulated capabilities to the next market after breaking through in one market?
1. The development path of local suppliers from the commercial vehicle to the passenger vehicle market
1.1 Regulations promote local autonomous driving Tier 1 to take the lead in finding a breakthrough in the commercial vehicle market
For local autonomous driving Tier 1 suppliers, the start is difficult - when the company was first established, although the technical reserves were slowly established, it was difficult to break through the market. OEMs were more willing to choose Global Tier 1 whose solutions had been tested by the market. , few of them are willing to take risks and bet on startups.
Especially in a field with high safety requirements such as automobiles, once the product quality is not up to standard, the consequences may be very serious. After OEMs use a supplier's solution, they generally tend to stick to the previous solution unless there is a new supplier that is much more cost-effective or the original supplier's supply is unstable. Therefore, in the first few years, many local Tier 1 companies had difficulty finding a breakthrough.
This situation slowly began to change in 2018 - with the successive implementation of commercial vehicle regulations, many commercial vehicles must be equipped with some safety-related ADAS functions. Compared with the passenger car ADAS market, the commercial vehicle ADAS market was very small (perhaps less than 1/10 at the time). Global Tier1 did not have a strong layout, and the OEMs were relatively receptive to local manufacturers, so some domestic The manufacturer seized the opportunity and first implemented the technology in commercial vehicles, bringing cash flow to the company.
At the beginning, the price of ADAS orders for commercial vehicles was quite high, and the profits obtained from each order were relatively generous. It was relatively easy for suppliers to use their technology accumulation to provide early warning and other functions for commercial vehicles. Therefore, after the implementation of relevant regulations, a batch of The company started placing orders for commercial vehicles. At present, most of the commercial vehicle ADAS pre-installation market is occupied by local suppliers.
1.2 After accumulating mass production capabilities, test the water on a small scale in the passenger car market
After successful cases in the commercial vehicle market, local Tier 1 slowly expanded into the passenger vehicle market, and new car-making forces became a breakthrough.
Compared with traditional OEMs, new car-making forces are smaller in scale and do not have much expense to support their search for large Tier 1 services at the beginning, because large Tier 1s generally tend to accept large-scale orders, and it is easier to do so if the scale is large enough. Dilute costs and generate sufficient profits. In the initial stage, it is difficult for new car-making forces to get strong support from international Tier 1 companies. However, when cooperating with local start-up Tier 1 companies that lack orders, they will have more "bargaining power" and receive better services.
Moreover, when local Tier 1 has implemented cases in the commercial vehicle market, everyone has some recognition of their mass production capabilities. Although commercial vehicles and passenger cars are not exactly the same market, their implementation experience in the commercial vehicle market is also To a certain extent, the engineering capabilities of local suppliers have been verified. The good service attitude and high degree of cooperation of local suppliers also made the project progress more smoothly.
In addition, at that time, the new forces still lacked brand effect and were not recognized by consumers. The cost of the models they produced needed to be controlled at a very, very low level to create the ultimate cost performance. Some local suppliers did a good job in cost control. Therefore, for new car-making forces that are just starting out, local Tier1 is a good choice.
On the other hand, for local Tier 1 startups, the most important thing is to start the project first, accumulate experience, and gradually iterate and improve. The R&D executive of Jimu Intelligent told the author: "We initially hope to have application cases, no matter how big or small it is. We need to start running first. It will be difficult for us to get orders from large OEMs at the beginning." Therefore, despite building new cars, The order scale of the powerful forces is relatively small, and the local Tier 1 will not "dislike" it.
Therefore, the new car-making forces at the beginning and the local Tier 1 need each other. Local suppliers first break through the small-scale orders of these new forces, and after polishing their engineering experience and proving their mass production capabilities, they will gradually "conquer" large OEMs.
1.3 Faster iteration speed helps local suppliers further open up the market
In the current era when product competition is particularly fierce, and OEMs launch new models almost every year, only a fast iteration speed can ensure the company's products maintain continuous competitiveness.
Usually, it takes about 3 years for Global Tier1 to do a mass production project, while local suppliers bring Internet genes to the field of automotive electronics, and it usually takes more than one year for a project to go from project approval to delivery. Then in terms of iteration speed, local suppliers have a great advantage.
Gradually, local suppliers have become a force that cannot be ignored in the passenger car ADAS market, and with the rapid increase of ADAS penetration rate, this momentum will only increase.
Today, many OEMs are cooperating with local suppliers to build assisted driving systems. There are also some local suppliers that have received strategic investment from OEMs. On the one hand, this means that OEMs are deploying industrial capital. On the other hand, it also means deeper cooperation between the two parties in the technical field.
The technology has been recognized, projects have begun to be implemented, and with the support of capital, local suppliers have rapidly expanded their scale. In the past year or so, the number of employees at many suppliers has doubled. After receiving many orders in succession, a certain supplier expanded its staff from 500 more than a year ago to nearly 1,200 now.
2. The consistent framework of commercial vehicles and passenger vehicles makes the migration of suppliers a logical process
2.1 Different companies’ plans for commercial vehicle and passenger vehicle business
From commercial vehicles to passenger vehicles, some local Tier 1 companies take advantage of the trend, first have a platform-based technology and product layout for the automotive OEM market, and then implement the functions that can be implemented first according to the actual domestic market conditions , such as smart driving technology MAXIEYE.
Since its establishment, Zhijia Technology's technology and product layout has been cross-platform for the automotive pre-installation market. After the commercial vehicle regulations began to be implemented in 2018, we obtained the commercial vehicle ADAS mass production project through this opportunity. According to the plan, once the time is ripe, more forward-looking or high-end technologies will be prioritized for large-scale mass production in the passenger car market. .
Of course, the terminal demand of the passenger car market lies in intelligent, personalized and experiential content. The figure below shows Zhijia Technology’s understanding of the differences between the two end markets of commercial passenger vehicles.
There are also companies that have clearly planned to open up the market with commercial vehicles as the first curve, and passenger vehicles as the second curve to expand the scale, such as Neusoft Reach.
Neusoft Ruichi started commercial vehicle ADAS mass production projects in 2016 and currently occupies a large share of the commercial vehicle ADAS market. Tu Jia, product planning director of Neusoft Ruichi's autonomous driving business line, said that the company's mass production capabilities have been greatly improved through a large number of commercial vehicle projects, especially in domain controllers. Domain controllers are a relatively difficult product in autonomous driving. When it later entered the passenger car market, Neusoft Reach absorbed a lot of its previous experience in mass-producing domain controllers in the commercial vehicle market. Neusoft Ruichi is opening up the passenger car market, mainly from two points of entry; one is the parking integrated domain controller, which was first mass-produced for Baojun Kiwi, and the other is the 5R1V and below solution.
After slowly opening up the passenger car market, many local suppliers have invested heavily in the passenger car market, actively expanded, and developed new technologies while maintaining their share in the commercial vehicle market, because the passenger car market is larger than the commercial vehicle market. 3-4 times.
2.2 Although there are differences between commercial vehicles and passenger vehicles, they have the same framework
The reason why these companies can do this is that although the commercial vehicle and passenger vehicle projects are not exactly the same, there are still many common technologies. Jimu Intelligent VP Dr. Wang Shuliang told the author: "We have been working hard in the field of commercial vehicles for so many years and have accumulated a certain amount of experience. Going into the passenger car market is more of a migration of our original capabilities."
The overall development framework for commercial vehicles and passenger cars is consistent, and the engineering experience honed in the field of commercial vehicles has also laid the foundation for the mass production of passenger cars.
3. Feasibility and challenges of migrating commercial vehicle ADAS technology to passenger vehicles
The consistent development framework provides the feasibility for manufacturers to migrate commercial vehicle ADAS technology to passenger vehicles, but due to differences in specific technical details, there will be some challenges in the migration process. So, which capabilities accumulated by manufacturers in the commercial vehicle field can be reused, and which ones need to be adjusted accordingly?
3.1 Reusable capabilities
3.1.1 Perception of reusable areas
Commercial vehicles have relatively large load capacity and large pitching and turbulence, and they often drive on national and provincial highways and other roads with worse working conditions. This means that the perception system of commercial vehicles faces much greater challenges than that of passenger cars. Therefore, as long as the accuracy of commercial vehicle perception is solved, it will be relatively easy to apply the perception solution to the passenger car platform.
At the data level, data collected in some generalized low-frequency scenarios can be reused, such as data for making qualitative judgments on target attributes and target cut-in intentions. In addition, when there is insufficient data, some manufacturers will fine-tune part of the commercial vehicle data and mix it with the passenger vehicle data for training. Of course, this is when the accuracy requirements are not particularly high and the scenarios are similar. .
At the algorithm level, Mr. Cao of Universal Intelligence mentioned that the basic network architecture is the same. For example, the recently popular visual BEV can adopt this architecture whether it is a commercial vehicle or a passenger vehicle. However, there are differences in different models and commercial and passenger vehicle scenarios. , the corresponding data set is needed to train and drive the algorithm model.
3.1.2 Where decisions can be reused
The decision-making level mainly involves model training. When manufacturers are working on commercial vehicle projects, they build a pipeline from accumulating data to training models, and how to accelerate this set of research and development processes on embedded platforms. These experiences are used in passenger cars. It can be used as a reference when working on projects.
3.1.3 Where execution can be reused
Compared with commercial vehicles, the body of a passenger car is lighter, and its body mass is relatively fixed. It does not have a large difference in body mass when the load of a commercial vehicle is different. Secondly, the chassis performance of passenger cars is usually better than that of commercial vehicles, so debugging will be simpler than that of commercial vehicles. Therefore, after making a good implementation plan for commercial vehicles, then starting a passenger car project is almost a "dimensionality reduction attack."
3.2 Places that need adjustment
3.2.1 Perceive areas that need adjustment
The heights and angles of sensors in passenger cars and commercial vehicles are different. Passenger car cameras are basically installed horizontally, with an installation height of roughly between 1.2 meters and 1.6 meters. Commercial vehicle cameras are basically facing downward, with an installation height of It is roughly between 1.6 meters and 2.8 meters, so the collected data will be different and cannot be completely reused. In addition, data related to accuracy basically need to be collected again, such as data used for quantitative target ranging and speed measurement.
Wu Guosuzhou, CTO of Zhenshi Technology, told the author, "If you directly use the data of commercial vehicles to train passenger cars, the effect will definitely become worse. The effect may not be obvious in most cases, but at close range or at large angles, for example In terms of vehicles, pedestrians, etc., due to differences in sensory data, the detection accuracy will drop a lot. This is unacceptable to users. Therefore, although the sensory data will have a certain similarity, it basically needs to be re-collected and annotated. "
Due to the difference in data, the adaptation scenarios are also different. When doing algorithm migration, some targeted tuning is required, and some branch modules may be added, or some tasks may be added to the deep learning model.
3.2.2 Where decisions need to be adjusted
When the same set of decision-making algorithms is applied to different models, it must be recalibrated. Because the body parameters of different models are different, such as tires, steel structures, engine positions, etc. may be different.
3.2.3 Implementation needs to be adjusted
The maximum deceleration of commercial vehicles is generally -5m/s², while the maximum deceleration of passenger cars can reach -10m/s². Because the body of a commercial vehicle is relatively high and may be a trailer, if the braking deceleration is too large, it may roll over. Then the timing for the passenger car to trigger braking can be later, and the parameters of the algorithm need to be adjusted accordingly.
4. Plan well and improve the organizational structure to facilitate the company to fully reuse the accumulated capabilities.
At present, Tier 1 manufacturers usually adopt a project system and arrange corresponding personnel from the company to these projects according to the specific needs of the project. For a company, there are many technical similarities between commercial vehicles and passenger vehicles that can be directly reused. To make full use of these commonalities to achieve technology reuse and resource saving, it is necessary to platformize the organizational structure. Make the overall design modular. The better the platform is done, the easier it is to reuse technology, the higher the overall efficiency of the company, and the easier it is to manage.
The head of R&D at Qingzhi Technology told the author that platformization and hierarchical management should be carried out as early as possible based on the V model of automobile development. Otherwise, many technologies that could have been reused cannot be reused, and they will be forced to do a lot of repetitive work, resulting in Waste of resources.
Specifically, in terms of hardware, the chip platform can be unified, so that the corresponding tool chain is the same set, and the algorithms and some interfaces that come with the chip can be reused. For internal development, this means that reinventing the wheel can be avoided.
In terms of software, we must do a good job in layering, and provide external communication interfaces, drivers, etc. for the underlying software. For the application layer software, we must segment the model into parts, such as perception and fusion, and then let the engineering team work according to customer needs. Make some tweaks. If this is not done well, and the layers and models are not divided properly, many bottom layers and application layers may be mixed together, which will be very troublesome when adjusting.
Correspondingly, there will also be some platform departments in the company's management structure. For example, in the area of perception algorithms, these technical personnel do not need to be divided into commercial vehicle projects or passenger car projects, and functional safety can also be shared; but it involves more differences, such as the front-end for different The product personnel and project personnel of the vehicle model, such as the system department, FAE, etc., and BD personnel need to flexibly change according to the project. People involved in specific projects often do not cross paths because they do not understand each other's respective needs.
unfinished words
Back to the essence, commercial vehicles are a production tool, focusing on safety and low cost; while passenger cars are a consumer product, within an affordable price range, the most important thing is to provide a good user experience. In terms of general driving functions such as TJA\LKA, commercial vehicles and passenger vehicles are the same, but there are many differences in other details.
Commercial vehicles, especially heavy trucks, as a means of production, are designed to help users reduce costs as much as possible while ensuring safety, minimize human fatigue during driving, or reduce the number of participants and reduce fuel consumption. , making the TCO of the product, that is, the lowest cost of use throughout the life cycle.
So when designing the upper functions of heavy trucks, how to reduce fuel consumption as much as possible is a very important topic, because fuel consumption costs account for a high proportion of the entire life cycle cost of heavy trucks, so design an algorithm so that the vehicle is equipped with automatic After the driving plan is completed, you can try your best to imitate the usage habits of experienced drivers, reduce fuel consumption, and create considerable economic value. Therefore, this is a very important topic for heavy truck projects. Many self-driving truck companies that focus on trunk logistics scenarios have regarded reducing fuel consumption as an important research project.
Passenger cars, especially those designed by many new forces, consider how to improve the user experience and create features that are easy for users to perceive and differentiate among a number of manufacturers, so as to facilitate the creation of selling points.
For example, some OEMs focus on improving the parking function and providing users with automatic parking services to make car use more convenient; some OEMs focus on creating higher-level autonomous driving functions so that users can be the first to experience the cutting-edge technology. science and technology.
With the rise of local autonomous driving supply chains and the continued emergence of economies of scale, the cost of autonomous driving solutions has begun to decline. For example, the cost of the simplest 1R1V solution has dropped to less than 1,000 yuan, and the price of lidar has dropped from hundreds of thousands of yuan at the beginning to several thousand yuan today. Now OEMs can create the ultimate solution for users with a very cost-effective solution. Intelligent experience. After overtaking in the field of new energy vehicles, domestic OEMs are actively deploying intelligence, improving the technological sense of the vehicle, and striving to compete with traditional luxury car brands in terms of product strength.
Conclusion: After experiencing the initial accumulation of technology and struggling to break through the market occupied by foreign capital, the scale of local autonomous driving suppliers has become considerable, and their competitiveness cannot be underestimated. The gradually growing local supply chain provides fertile soil for the development of smart cars in China, and the development of the smart car market will feed back the supply chain. I believe that as time goes by and technology continues to iterate, China’s smart car market will develop better and better.
write at the end
Communicate with the author
If you want to communicate directly with the author of the article, you can directly scan the QR code on the right and add the author's WeChat account.
Note: Be sure to note your real name, company, and current position when adding WeChat
As well as information such as intended positions, thank you!
About submission
If you are interested in contributing to "Nine Chapters of Intelligent Driving" ("knowledge accumulation and compilation" type articles), please scan the QR code on the right and add the staff WeChat.
Note: Be sure to note your real name, company, and current position when adding WeChat
As well as information such as intended positions, thank you!
Quality requirements for "knowledge accumulation" manuscripts:
A: The information density is higher than most reports of most brokerage firms, and not lower than the average level of "Nine Chapters Smart Driving";
B: Information must be highly scarce, and more than 80% of the information must not be seen in other media. If it is based on public information, it must have a particularly powerful and exclusive perspective. Thank you for your understanding and support.
Recommended reading:
◆ Nine Chapters - A collection of articles in 2021
◆Detailed explanation of Inceptio’s self-driving truck mass production technology methodology
◆ Lidar: The war between 905 and 1550
◆ The single SoC chip solution may accelerate the large-scale mass production application of the integrated travel and parking solution