Editor's note: "Software-defined" is a new feature and new symbol of a new round of technological change and industrial change, and has become an important force driving future development. From now on, this newspaper will launch a series of reports on "Software-Defined World", focusing on the innovative vitality and vivid practice of "software-defined" in the automotive, industrial and networking fields. Stay tuned.
At the beginning of 2022, the biggest merger and acquisition in the automotive industry occurred in the software field, with the ACQUISITION of Wind River Software by US auto parts giant Aptiv for $4.3 billion. Recently, the Volkswagen Group of Germany shifted the focus of its R&D process and organizational management from components to functional software for vehicle systems. And Tesla's rise is also due to its bold innovation in software...
When traditional cars gradually evolve to mobile smart terminals, car companies suddenly realize that selling hardware has become less and less profitable, and software has become the biggest variable in the new law of competition. Whether it is traditional car manufacturers or later "new forces", they are increasingly betting their business models on their own software strength.
Software controls the evolutionary power of the car
10 years ago, software was just a simple real-time program embedded in the car, deeply coupled with the ECU (Electronic Control Unit), controlling an 8-bit or 16-bit ECU, performing a relatively single function. Driven by the wave of intelligence and networking, the ongoing demand for functional evolution and security patching requires automotive software to be updated at least every three months. As a result, automotive software and hardware are separated at the component level – decoupling hardware that tends to be standardized and software that can grasp the evolutionary capabilities of the car.
Tesla even turned the car into a "computer." Technicians who have disassembled Model3 revealed that Tesla has realized the separation of software and hardware, merging dozens of ECUs into three computing units, and one software can call multiple hardware, realizing many new functions.
In February this year, due to the abnormal opening of one of the valves in the heat pump, resulting in a decline in the defrosting capacity of the windshield, Tesla recalled the vehicle involved in China, through the OTA (space download technology), the function of the camera and sensor where the problem was located was optimized, and the whole process took only a few minutes. Previous similar recalls, users had to go to the car dealership's store to implement manual software updates or replace hardware.
"Automotive hardware determines the lower limit of a car's intelligence, while automotive software determines the upper limit of a car's intelligence, and undertakes the user's most intuitive intelligent experience." Wu Longba, co-founder of Skyworth Automobile, told the Reporter of China Electronic News. From the basic control of the vehicle, to the advanced function of intelligent cockpit software, car networking software, automatic driving... Software has become an independent core component, gradually gaining full-stack and complete control permissions.
Tesla's design language and interactive experience innovations are entirely software-based thinking. In 2012, Tesla introduced OTA to its own ModelS for the first time in the industry. By the end of 2021, Tesla has carried out no less than 70 OTA function upgrades, in addition to traditional functions such as applications, maps, lights, etc., it has also upgraded the relevant functions of intelligent driving systems such as forward collision warning system, lane departure warning system, automatic emergency braking system, and parallel assist. While repairing past defects and improving performance, feedback analysis is carried out according to customer data to further accelerate product iteration speed and enhance product retention rate.
FSD (Full Self-Drive) is arguably one of the most important pieces of software for Tesla cars, mastering the evolutionary direction of future autonomous driving. A Tesla user who purchased the FSD software package said that the optional FSD bisu car has more automatic assisted driving, automatic assist lane change, automatic parking and intelligent summoning functions, which is "a function that can never go back when used". Tesla CEO Elon Musk believes that with the FSD system, Tesla may completely abandon radar in the future.
SAIC motor is infiltrating "software thinking" into the process of car development, verification, sales, service, etc., and constantly changing and optimizing various processes. SAIC Zero Beam Technology Co., Ltd., which was established in November 2021, launched a new "Galaxy Intelligent Vehicle Full Stack Solution". Li Jun, CEO of SAIC ZeroHub, said that the new solution includes four basic technical solutions: centralized electronic architecture, SOA (service-oriented architecture) software platform, intelligent vehicle data factory, full-stack OTA and network security solution, and an end-to-end intelligent scenario service platform that brings users the ultimate experience.
After the separation of software and hardware, the central centralized electronic architecture of the car type realizes the pluggable, pullable and scalable hardware; the SOA software platform realizes the software can be bought, sold, and subscribed; and the data factory realizes the continuous self-learning, self-growth and self-evolution capabilities of smart cars. At present, SAIC Zero Beam has opened 680 car-side services, 980 cloud services and 260 data services for upper-level developers, covering more than 20 fields such as vehicle control, intelligent driving, infotainment, intelligent networking, and data training.
Or subvert the automotive supply chain system
More and more car companies have found that hardware has become less profitable, and software services, software updates, and software charges have gradually become the profit points of car companies.
GAC Weilai announced the cost of its parts, saying that the comprehensive net profit margin on hardware is not higher than 1%. "What materials are used by enterprises, how much money they need, what equipment they use, and how many processes they have passed, the cost calculated is very clear, and the value of software and ecology is the differentiation of smart cars." Liao Bing, CEO of GAC Weilai, said.
Under this trend, the automotive industry has gradually changed from the manufacturing industry of ToB to the modern service industry of ToC, which has also pushed down the traditional organizational model of many car companies.
Recently, the Volkswagen Group of Germany has initiated the adjustment of the company's technology research and development structure, shifting the focus of the R&D process and organizational management from components to vehicle system functions. Volkswagen said that when the basic software architecture is in place, the vehicle project cycle will be shortened from 54 months to 40 months, and the future delivery to customers will be cars that can be updated in real time.
With the internal organizational structure of car companies shifting from "hardware first" to "software first", the hardware-led automotive supply chain system is being subverted, and the original supply partnership may gradually become a competitive relationship.
A person in charge of an auto parts company told The China Electronics News that if there were originally 60 ECUs on a car, which came from different first-tier suppliers, when the electrical and electronic architecture was concentrated on 4 domain controllers, it meant that some suppliers would disappear. Some suppliers, who originally played the role of integrators, have to hand over some control and readjust the cooperation model with customers.
Following the footsteps of upstream parts manufacturers seeking downward extension, parts companies have also begun to build "soft power", link traditional businesses, and increase the size of high value-added fields.
Auto parts giant Bosch has established the Intelligent Driving and Control Division to provide electronic systems and essential software to existing and new customers. Bosch said that it can provide only software, or only hardware, or it can also provide software and hardware integration solutions. They do not think that OEMs can be fully responsible for software research and development independently, and this assumption cannot achieve the "optimal solution" in terms of resource investment and development efficiency.
Continental, another component company, invested in an AI (artificial intelligence) supercomputer to meet the speed of deep learning, simulation and virtual data generation of autonomous driving systems, and accelerate the commercialization of unmanned driving systems.
At the same time, the boundaries between automotive and other industries are blurring. "Internet and ICT companies with software research and development advantages are seriously squeezing the original supplier space, and gradually forming an intelligent network supply system dominated by who is strong." Wei Li, senior director of electronics and electrical at the China FAW Intelligent Network Development Institute, told China Electronics News.
In 2019, Huawei entered the automotive company as a "full-stack supplier of smart car software and hardware"; the following year, Huawei demonstrated a set of software and hardware solutions and cooperated with changan automobile, BAIC BJEV and other car companies to develop new models. In 2021, Huawei Hongmeng built a third-party platform for all fields and began to help car companies improve the cross-platform capabilities of their operating systems.
Baidu is the earliest layout of intelligent driving in China's leading Internet enterprises, in 2013 on the basis of the deep learning research institute set up an automatic driving research team, in 2017 released the automatic driving software Apollo 1.0, has formed automatic driving, vehicle-road collaboration, intelligent vehicle connection and other three major open platforms, to help Chery, Great Wall, Ford and other car companies combine vehicle hardware systems to build their own automatic driving systems.
Automotive companies are building software barriers
The mainstream consensus is to divide automotive software into three categories: the underlying operating system, the tool software and the application software. Although cross-border cooperation often occurs in the automotive industry, out of the consideration of mastering core functions and ecological chains, automakers often build technical barriers through self-developed software to seize the right to speak.
"For car companies, the application layer software must be self-developed, and the middle layer software is best mastered by themselves." Wu Xuebin, deputy general manager of Chery Automobile Co., Ltd. and general manager of Xiongshi Technology, pointed out in an interview with China Electronics News that "the underlying software and hardware (chips) are strongly related, and automobile companies do not need to make chips, but they must have a very good understanding of chips and make a perfect combination of software and hardware." ”
The automotive software market is growing in size
Source: CCID Consultants
CCID Consultants and Neusoft Group jointly released the "Insights on the Development Trend of China's Intelligent Vehicle Software Industry", which pointed out that the current growth of China's intelligent car software industry is mainly based on application software, of which automatic driving software is the main increment, which has become the focus of competition for automobile companies. Large enterprises related to the upstream and downstream of the industrial chain mainly carry out strategic layout in the field of intelligent vehicle software through direct participation, capital injection and establishment of subsidiaries and acquisitions.
Tesla's self-developed self-driving FSD software sells a quarter of the lowest price of the Model 3, and it is currently the world's highest-priced vehicle automatic driving assistance product. Musk said that when FSD becomes more and more feature-rich, the price cap may break through $100,000. Tesla CFO Zach Corkhorn said on the Q1 2021 results call that FSD revenue will reach $12.9 billion and gross profit will reach $9.7 billion in 2025. Essence Securities speculates that in 2025, FSD will contribute 25% of the gross profit of Tesla's automotive business.
Li Pengcheng, vice president of Xiaopeng Automobile, told China Electronics News that intelligent assisted driving is the differentiated competitive advantage of Xiaopeng Automobile, so it adheres to "full-stack self-research" in this field to form a full closed loop of data and algorithms.
Xiaopeng Automobile not only develops the visual perception, sensor fusion, planning, positioning, decision-making, control and other aspects of the vehicle side, but also realizes self-research in the data upload channel, front-end data upload implementation, cloud data management system, distributed network training, and data collection tool development required for cloud data operation. According to Li Pengcheng, after purchasing a model with XPILOT 3.5 hardware, Xiaopeng Automobile users can choose whether to purchase XPILOT 3.5 software, increase the intelligent driving assistance functions of urban roads and highways, and high-level intelligent driving assistance functions such as cross-floor memory parking functions. The revenue of XPILOT 3.0 (high speed) and XPILOT 3.5 intelligent driving assistance software services has also become an important part of the company's revenue.
"The assisted driving systems of many new car companies are comparable to Tesla's. But the real difference is the combination of hard and soft. Self-developed car chips to help Tesla's FSD as much as possible to use computing power, with the iteration of autonomous driving chips and algorithms, even if the software is continuously upgraded, fluency and computing power are enough, and there is no need to purchase additional chips with the upgrade, and the development cost of the whole vehicle will be greatly reduced. Shao Yuanjun, research director of the Chinese Academy of Sciences Venture Capital Management Co., Ltd., told the China Electronics News reporter.
Traditional car companies are also moving. In foreign countries, large traditional car companies represented by GM and Ford have set up new companies such as Cruise and Argo AI through capital injection, and conducted layout research and development in L3-L5 level automatic driving software algorithms. In China, Great Wall Motors established Zhixing, and SAIC Motor established a software center SAIC Zero Beam to develop autonomous driving systems with a data-driven end-to-end development process.
In addition to application software, self-developed operating systems have become a new track for car companies to compete for deer because they can help simplify the vehicle software development process and increase the frequency of OTA updates. Earlier this year, Toyota motor announced plans to launch its self-developed Arene operating system in 2025, and Mercedes-Benz is expected to release its self-developed MB.OS operating system in 2024... Domestic and foreign automotive leading enterprises are becoming more and more demanding to achieve autonomous controllability and unification of the on-board operating system.
The route of car companies in the field of automotive operating systems is also gradually becoming clear. One is based on Android, Linux and other underlying system architecture for in-depth customized development, including system kernel modifications, such as Volkswagen Group's VW.OS, Tesla Version, Huawei Hongmeng OS, AliOS, etc.; the other is based on Android, Linux, etc. for limited customized development, but does not involve system kernel modifications, such as BYD DiLink, Geely GKUI, NIOS, Xiaopeng XmartOS, etc.
The operating system helps car companies bring direct benefits by enveloping the ecosystem. Relying on its annual sales of nearly 10 million cars, the Volkswagen Group forced Tier1 (a tier 1 supplier that signed a supply contract with the host manufacturer) and software suppliers and even other OEMs to develop on the basis of VW.OS, making it an "iOS in the smartphone field", and finally forming a business model of "OS licensing fee + Car Networking service + APP docking license fee + APP value-added service share", so that its operating profit in 2021 will double in the case of declining sales. It surged 99.2% year-on-year.
Don't let software become the "soft underbelly" of the car
On the one hand, the rapid development of smart cars, on the other hand, the car function runs on top of the software, and the safety hazard surface will also increase.
According to data from the State Administration for Market Regulation, in 2021, 2,423 automotive product recalls were implemented in China, involving 91.3 million defective products. In terms of the recalled models of the top ten car brands/car companies in 2021, the reasons have extended from traditional hardware failures to more and more system program and software design problems.
According to the statistics of the 360 Vehicle Networking Security Laboratory, among the 53 models of 25 domestic car companies that are selling intelligent networked cars, 360 companies have found a total of more than 1600 vulnerabilities, of which more than 1000 cloud vulnerabilities can cause attackers to remotely control all the intelligent networked cars of the brand; more than 600 car-side vulnerabilities can lead to close non-contact control of cars, such as opening and closing doors, starting engines, etc.
"Software-defined cars make security breaches inevitable." Zhou Hongyi, founder of 360 Group, said that there are 4-6 vulnerabilities per thousand lines of code on average, and troubleshooting vulnerabilities will be the "new normal" that accompanies the whole life cycle of intelligent networked vehicles from production to destruction.
At the same time, the physical security of the car is upgraded to digital security, and the advanced network threats against the vehicle software and hardware, communications, cloud platforms, data, supply chain, APP, etc. will become the primary security problem of the car, and the attack means will break through the latest styles including supply chain attacks, ransomware attacks, firmware attacks, etc. The attack effect can directly lead to the destruction of the car.
Zhou Hongyi suggested improving the automotive software security index from three aspects: one is to guide network security companies to discover and deal with automotive security vulnerabilities in accordance with laws and regulations; the other is to establish a long-term mechanism for "digital space crash test" of intelligent networked vehicles as soon as possible, and conduct network penetration tests on on-board software and hardware, communications, cloud platforms, data, supply chains, apps, etc. through professional and standardized network attack means to test the network security and data security capabilities of the whole system of the "cloud, tube, and end" of the tested car The third is to build a digital security capability system for intelligent and networked vehicles with the automotive safety brain as the core.
The more advanced, the faster, and the more vulnerable. Software has risen to the height of defining the function of the car, and the construction of the security protection system must keep up with the speed of iterative evolution of software, and do not let the software "armor" become the "soft underbelly" of the car.