Translation Result

Title: On Chang’an Automobile’s Technological Transformation

The changes in Chang’an came somewhat unexpectedly.

At last year’s “Chang’an Automobile Technology Ecological Conference”, the keyword for Chang’an was still “Avita”. Together with Huawei and Ningde Times, the industry’s focus was on the product level, with doubts about traditional host factories’ desire to transform into technology companies and a vague understanding of SDA’s intelligent digital architecture, which is Avita’s parent company.

At this year’s conference, all of the above questions became clear.

Image here

If the goal of transformation for auto companies is not to become technology enterprises, it would be directionally wrong. However, technology is a broad concept, and auto companies generally need several basic points, such as highly integrated design (such as the integration of the three-electric system, chassis and battery, etc.), weaving digital technology, supporting algorithms and visual perception technology to achieve higher-order automatic driving, and a “digital platform architecture” that governs the above items.

At the “Second Chang’an Automobile Technology Ecological Conference,” although the path shown by Chang’an Automobile may not be the standard answer to the transformation of auto companies, it makes it clear to the market what Chang’an Automobile has accomplished and can deliver to users.

Achievements in Phases

In phase achievements, Avita 11 has received 20,000 orders, the recently released Deep Blue SL03 has received over 30,000 big order, and Chang’an Lumin has received over 30,000 orders in its first month.

Data is not a direct reflection of technology, but in the high-end pure electric vehicle market where “no intelligence is left out,” without two master skills as product labels, sales data will certainly not be too pleasing to the eye.

Chang’an has provided several single-item achievements, such as Super Set Electric Drive, Electric Drive High-Frequency Pulse Heating, iBC Digital Battery, Chang’an Wisdom Core, and Intelligent Parking Technology, among others. The specific interpretation is covered in previous articles (“The End of the Evolution of Auto Companies Might Not Be Technology Companies”), and will not be further explained here.

Image here

However, the distribution of technologies is worth noting, involving structural design, thermal management, domain controllers, and autonomous driving, and it is not just a futures item. The SL03 that is already on the market is already equipped with them. This is a microcosm of the skill tree of host factories during the transformation process. In the past, tasks that could be handed over to suppliers to solve, are now the bottom line of “you must know a little more.” There is a trend toward full-stack self-research, which also outlines the embryonic form of future tech companies.

On the product platform side, before the appearance of SDA, Chang’an’s official stance was that it had two platforms in hand:- One is the intelligent electric-connected vehicle platform CHN, which cooperates with Huawei and Ningde era. According to Changan’s definition of Aweta 11, this product is positioned as a “companion role”, with emotional experts’ capability.

• The other is the full-electric platform EPA1, which is compatible with three types of energy sources: pure electric, extended range, and hydrogen fuel. It is the mother platform of SL03.
• From the perspective of platform significance, both emotional experts and full-electric power experts do not exceed our understanding of traditional automotive platforms. Especially for the latter, its DNA is coverage and breadth, and there is still no sign of “breaking away” from the petal-shaped horizontal extension of the fuel car platform.
• What really needs us to re-examine the “automotive” platform (let’s call it that for now, maybe it will not be accurate in the future) is the SDA architecture that has landed with CD701.
• Changan divides the SDA architecture into six layers, from body, energy, to brain, operating system, external application, and beyond body cloud big data.
• An intuitive understanding is from “flesh” to spirit, such as L1 being the bottom layer, which can complete some difficult mechanical missions, such as adjustable chassis height and suspension adaptation, and still adopt high-integration design.
• The upper layer of the mechanical layer is the energy layer, which, as the name suggests, is the source of power. The numbers are also impressive, such as the platform product can break 0-60 mph in 3.8 seconds, the range is 900 kilometers based on CLTC standards, and under the ultra-fast charging mode, it can be charged for 150 kilometers in five minutes.
• The benefits brought by the layered architecture are software-hardware decoupling and software-software decoupling (bottom-level software and application software).
• For example, Changan describes CD701 born on the SDA architecture as “through plug-and-play upgrades… changing chips without changing cars, using new cars at any time… equivalent to having multiple device hosts such as notebooks, game consoles, etc.” such general IT terms.Leaving aside the tags that store game consoles, this expression has two meanings: first, powerful performance, wide application; second, product upgrades are not limited to OTA, but hardware upgrades can be completed in plug-and-play form, without sacrificing the entire vehicle.

This is not only similar to the past computer upgrades, which only required replacing the processor, but also means that in the future, software and hardware can be upgraded separately.

This is based on the centralized electronic architecture of SDA, which can unlock intelligent driving, multi-screen interaction, game-level 3D effects, and achieve “changing the core without changing the car” in terms of hardware. This decoupling of software and hardware has changed the cumbersome process and half the effort of adding an electronic function to a traditional fuel vehicle, which usually requires adding an ECU. Software updates are not limited by hardware, which means that the speed of “car” updates and upgrades is accelerating, such as the traditional automobile product release cycle generally taking several years, but in the future, the update cycle may only need a few months.

On the other hand, the integrated design simplifies the number of ECUs, making it easy to integrate them into domain controllers while reducing costs. With the advancement of platformization, hardware expansion and consumer electronics access will also become a reality, and plug-and-play hardware upgrades will also lay the foundation for automatic driving upgrades.

At the same time, the underlying software and application software have also been decoupled. Chang’an stated that this service-oriented software architecture has achieved 1000 services and interfaces that can be debugged, directly copied and transplanted, and supports the connection and control of tens of millions of intelligent vehicles.

The Arrival of Intelligent Robots

In the 1990s, James Cameron’s “Terminator” formed our initial idea of ​​future machines, but at that time, the T800 had two limitations: it could not be upgraded online, and it could not interact with cloud data. If these two passageways had been cleared at the time, I believe the battle between him and the T1000 would not have been so tragic.

These two limitations have now been cleared. In fact, Chang’an also regards future cars as intelligent robots, which not only have learning consciousness, but also can interact with cloud data, allowing experience values ​​to be shared and thus iterated quickly.

This is just the tip of the iceberg for the L5 whole vehicle functional application layer and L6 cloud big data layer in SDA architecture, presented in an open-ended manner because there is a lot of room for imagination in future growth. For example, the former belongs to the skill unit of the product, while the latter belongs to the knowledge unit, but the former can also be original in skills as it can be conformed to the cloud big data that can handle over 55PB of processing power, with a GPU computing power of 45 PFLOPS and a daily processing capacity of up to 10 billion data through atomic services and programming (the underlying software and application software are separate, and the hardware can also be upgraded separately). As for what specific functions can be created, we can imagine for ourselves.

The L6 cloud big data layer itself is a closed loop of data integration between the cloud and the end device, belonging to the robot’s knowledge unit. Changan has divided this thinking generation logic into several steps, such as “collection, labeling, training, simulation, and calculation.” Sharing is the essence, and subsequent digestion and absorption are efficiency advantages.

This also creates expectations for the evolution of its products in perception, decision-making, and behavior, and is one of the core elements of Changan’s “Zhuge Intelligent” brand released this time.

Cloud computing power carries massive information processing, and information supports algorithm evolution. For example, the interactive strategy brain introduced by Changan currently pursues the ability to achieve close to natural perception in vision, speech, and comprehensive biological recognition engines, as well as a comprehensive coverage from humans to vehicles in behavioral (including drivers), vehicle and environmental data analysis engines in terms of intelligent analysis dimensions.

From cloud interaction to upgradable SDA architecture in layers, it actually corresponds to the needs of users at different levels. This is the biggest advantage of decoupling. Not everyone is obsessed with high-level autonomous driving, and Changan does not reject users who prefer active control over products.

So it’s all about individual preferences. For example, the economic computing platform can connect to 22 sensors to achieve adaptive cruising, highway assistance, remote smart parking, and other “basic operations”; the medium computing platform supports 36 sensors to achieve high-speed navigation assistance, remote customer parking, etc.; the high computing power can connect to 48 sensors to achieve intelligent driving in all scenarios, with computing power of up to 508 TOPS.

This actually breaks through the rough predictions of some industry insiders on the prospects of autonomous driving in the past, either adopting or rejecting it entirely. This is where the automotive industry is making progress, and it has achieved a rating system before the film industry.

Conclusion

There is still a lot of knowledge to be shared, but we are concerned that readers may get fatigued. For instance, the CD701 car actually applies the concept of a living space and also has strong expandability, with 1,000 services and interfaces that can be debugged, making it one of the highlights of this car’s software architecture.The image of the technology industry in most people’s minds is probably still based on the image of the rocket built by Tesla launching into space. However, at present, some domestic enterprises have taken the lead in integrating concepts similar to Tesla, such as the integrated concept, which is not only used in battery chassis integration but also reflected in the electronic and electrical architecture.

Compared with other enterprises that use products to represent their behavior and vision, Changan has built a meticulous platform architecture system that is trying to run through an upgraded and evolved logic. It has added weight to both human-machine interaction and autonomous driving. It is also the leading automaker among many technology enterprises that are currently intending to transform with high credibility.

This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.