Beijing Jiuzhou Huahai Technology Co., Ltd. (referred to as Huahai Technology), founded in 2011, is a leading domestic provider of automotive electronic control system solutions. Relying on its independent core technologies, the company has profound accumulations in embedded systems, core three-electric systems control for NEVs (New Energy Vehicles), automatic transmissions, and fuel cell engine control, serving multiple industries including automotive, machinery, and marine. With high-quality products, advanced development tools, and comprehensive technical services, the company has won the trust of numerous clients such as SAIC, FAW, Dongfeng, and SANY Heavy Industry, with its products exported overseas.
As the competition in smart vehicles shifts from horsepower to computing power, a hidden battle over software efficiency has long been underway. Automakers, inspired by the vision of "software-defined vehicles," find themselves trapped in development delays and out-of-control costs. Traditional development models are reaching their limits, and the industry is in urgent need of an "efficiency revolution." This episode of "Chip Unveiled" will deeply analyze how automotive software engineering can achieve an order-of-magnitude improvement in development efficiency through the iteration of development paradigms.
Welcome to the dialogue podcast, *the following is the text transcript of the podcast, confirmed and authorized by the guests.
Is Code Trapping Veteran Engineers? Traditional Development Models Enter the Twilight
Huan Shi: In recent years, we have constantly heard a term—software-defined vehicles. Given the current situation where the cost of Electronic Control Units (ECUs) is relatively high and the development cycle is long, what do you think is the biggest development challenge for ECUs? Because I know your company has been deeply engaged in this field.
ECUs are responsible for receiving signals from various sensors across the vehicle (such as speed, temperature, and position sensors), performing calculations and processing based on internally preset programs and data, and ultimately outputting commands to control actuators (such as fuel injectors and relays), thereby precisely controlling various operational states of the vehicle.
Ding Wenchao: With the evolution of NEVs and intelligence, a massive amount of software in vehicles needs to be customized, and software functions are becoming increasingly complex. Therefore, traditional distributed controllers struggle to meet current requirements. Originally, one function corresponded to one controller, but now there are too many functions. Over years of development, this has grown into hundreds of controllers, which makes software collaboration efficiency extremely low, as they may come from different suppliers.
In the so-called "software-defined vehicles," or the transition of vehicles from functional cars to smart cars, how to solve the complexity problem brought by software customization, especially in scenarios involving reliability and safety where software complexity requirements are even higher, poses a significant challenge to us. Huahai Technology has always focused on this field, developing software development processes and toolchains to enable software to be completed with 3 times the efficiency.
Huan Shi: Are you targeting traditional ECUs or new domain controllers? The technical differences between them should be quite significant. Could you please explain this to us?
Ding Wenchao: Distributed architecture is still the mainstream now, but domain control architecture also has considerable applications. We are currently collaborating with many automakers on development. Basically, for next-generation models, automakers generally choose the domain control architecture, including some more advanced ones like centrally controlled models, and there are also some concept cars being launched.
Following this trend, these directions are actually all covered under the development route of domain control. Then the collaboration between automakers and the supply chain is also crucial, especially the OEMs' understanding of software, which is equally vital.
Vehicle Control Unit (VCU) system (Image source: Huahai Technology)
Huan Shi: The middle layer, which is the software layer, is extremely important, and you need to integrate a lot of horizontal understanding. This is not easy, especially now that everyone is paying more and more attention to electronic control safety and the overall user experience of the vehicle. What are your feelings in this area? What key changes have occurred in customer demands?
Ding Wenchao: The functions and user experience in the vehicle are strongly correlated with software. To ensure the experience, two core aspects must be achieved.
First is software quality. In the past, we adopted the V-Model Development process. Under this model, the software quality of functional vehicles was significantly improved. However, entering the era of smart vehicles, the development cycle is becoming shorter and shorter. In the past, the biggest impacts on the vehicle development cycle were, first, the development of the electrical/electronic (E/E) architecture system, and second, the development and matching of systems like engines and transmissions.
V-Model Development: Refers to the most classic systems engineering methodology in the field of automotive embedded software. The left side decomposes the design, and the right side integrates and tests. Through layer-by-layer corresponding verification, it ensures development quality and functional safety. It is named after the letter "V" because the development process is graphically shaped like it.
Now in the NEV field, the impact of engine and transmission development no longer exists. Although there are development cycles for the newly emerged batteries and motors, the impact is not too significant. What truly brings pressure is the complexity of the E/E architecture and the complexity of software functions. Against this backdrop, the collaborative processes between both parties, including the software development model, must be further upgraded.
Second is requirements engineering. The past V-Model, at this current stage, has actually been "ruined" by us in some cases. Because the V-Model originally included three major engineering phases: requirements engineering, development engineering, and testing engineering. But the common phenomenon is that people do not pay attention to requirements engineering, resulting in the phase from development to testing, especially the testing phase, being dragged out longer and longer. Many people want to rely on testing to ensure software quality, but the result may be that the more they test, the more problems arise. Because the trade-offs at the requirements end might be wrong, and some requirements are not judged well or are defined problematically, testing will only continuously expose problems, leading to continuous patching. This situation is highly detrimental to software quality.
Huan Shi: Is this related to everyone's eagerness to update and iterate too quickly? Because I can clearly feel now that the cycle for each automaker to launch its own products is much shorter than the traditional five or eight-year cycle.
Ding Wenchao: Yes, this is also a problem we have to face in reality. It is equivalent to automobiles transforming from original mechanical products into electronic products of smart hardware. The iteration speed of smart hardware is a challenge to our traditional automotive reliability development process. Because if you want to ensure reliable quality, some key steps must be taken. If you need to ensure the schedule at this time, you have to use more agile methods for development.
Cao Huanshi, Founder of Chip Unveiled (right), in dialogue with Ding Wenchao, CEO of Huahai Technology (left)
Breaking the Efficiency Bottleneck: Graphical Tools Become the "New Favorite" of 90% of Engineers
Huan Shi: Indeed, many domestic automakers pay more attention to application scenarios and less to the underlying layers, which is the so-called "emphasizing applications while neglecting the underlying layers." What breakthroughs have you made in the underlying layers, such as in toolchains or processor adaptation? I see that the software and hardware companies you cooperate with are all international giants, like NXP and MathWorks, which are your important partners.
Ding Wenchao: Huahai Technology has over 20 years of accumulation in the operating system and complex driver levels of the powertrain domain. The powertrain domain has long been one of the most complex software areas in vehicles, and it is also an area with high requirements for real-time performance, safety, and reliability. Focusing on this field, we have built a graphical programming embedded real-time operating system, which we call "eCore."
With this operating system, our engineers can very conveniently implement the functions and configuration of the underlying software merely through some menu-based drag-and-drop operations. It is somewhat like we have created a Windows operating system for automotive controllers, which can help engineers run their various innovative ideas well on this embedded platform and implement control in the vehicle.
Huan Shi: Is the basic threshold for customers to experience this system high?
Ding Wenchao: It is very convenient for users to use.
The beginning of automotive software caught up with the Internet wave. A large number of practitioners transitioned from majors like vehicle engineering and power engineering to software development, unlike traditional Internet practitioners who were originally trained in computer science and software engineering. Therefore, at the programming level, the automotive industry indeed has a certain gap.
The graphical programming operating system we built adopts fully automatic code generation, allowing these systems engineers to develop in a zero-code manner. They do not need to know programming; they only need to clearly express their professional knowledge and logic. Because software is essentially a collection of knowledge and logic, our system simply converts it into code automatically. Through this graphical programming combined with automatic code generation, enterprises can learn it particularly fast. Basically, from design to the implementation of vehicle control functions, it takes as short as one month, and no more than half a year at the longest, which is highly efficient.
RapidECU operation interface (Image source: Huahai Technology)
Huan Shi: How did you implement the graphical aspect? Is it easy for customers to get started?
Ding Wenchao: The graphical aspect is actually developed based on the Simulink software environment from MathWorks. Our fully automatic code generation tool (ECUCoder) also needs to be used in conjunction with toolboxes like Simulink Coder and Embedded Coder from MathWorks. Moreover, this development method, after more than ten years of promotion by us, has become the mainstream choice in the industry. Now, basically 90% of automotive software engineers are using this method, forming an industry habit.
Our set of tools is relatively easy for customers to get started with. Once they get used to it, it is hard to go back to the traditional manual coding method. Because the traditional method is inefficient and very time-consuming and labor-intensive to use.
Huan Shi: Would platforms like DeepSeek, which can use AI to generate code, be a challenge to you?
Ding Wenchao: Currently in automotive software, AI is more of an enabler. In the past, programming required consulting a large number of manuals to call interfaces. Now, these operations can be automatically completed through AI, which actually greatly improves efficiency. At present, it is still hard for AI to replace us.
Combining Hardware and Software: New Strategies for Automotive Cost Reduction and the Future of AI Electronic Control
Huan Shi: As a service provider that works closely with OEMs, under the pressure of cost reduction that Chinese OEMs face every year, do you have any strategies to help customers reduce costs? For example, using your tools, or is there a possibility for your system platform to achieve cost reduction?
Ding Wenchao: Our products have two cost reduction advantages. First is the reduction in hardware unit cost. We will integrate supply chain resources, including cooperating with international-level contract manufacturers like Foxconn, utilizing their capacity and cost control capabilities to optimize hardware cost and reliability. Foxconn currently has a lot of idle capacity, and utilizing it can avoid the waste of repeated investments.
The automotive industry is highly scale-dependent. It might require manufacturing at a scale of 5 million or 10 million units to achieve a certain cost reduction goal. In the entire field of innovation, the initial volume might not be that large. Moreover, during the entire capacity ramp-up process, we found that the actual annual sales of NEVs are not that high. This requires more efficient ways to utilize more social resources and reduce costs from the perspective of hardware unit costs. This is one aspect.
The second cost reduction advantage is software cost. For intelligent connected NEVs, there are two major costs: the vehicle software and the battery. The cost of vehicle software may account for 25% or even over 40% of the total vehicle cost in the future. How to reduce software cost is a significant challenge for the entire current industry chain ecosystem. In this field, what we consider is greatly improving software development efficiency. Just as we said, we can increase software development efficiency by 3 times, and by utilizing our entire set of more advanced development models, we can reduce software development costs by 80%.
Huan Shi: In the next 3 to 5 years, do you think there is a possibility of achieving major breakthroughs in technology in the field of automotive electronic control? Have you made any technological layouts in this regard?
Ding Wenchao: The most important trend in electronic control now is still AI technology, which we in the industry call embedded control terminals, an application on such smart hardware. In the past, hardware was based on some control algorithms based on rules, which is a mode of defining rules in known scenarios and letting it run according to these rules.
However, with the application of AI, controllers can become smarter. They will have a certain degree of autonomous learning ability, as well as judgment, decision-making, and planning capabilities. In this way, electronic control can play a higher role in energy conservation and emission reduction. Especially in fields related to vehicle motion safety, energy consumption has always been a significant challenge for us. How to minimize energy consumption and maximize energy utilization efficiency, such as the utilization of various clean energy sources, are all problems we need to further solve through AI.
Ding Wenchao: In addition, various intelligent functional experiences, especially those involving vehicle driving performance, will have more personalized and customized experiences. These can also be supported by our new-generation E/E architecture platform foundation, making it more convenient to integrate various personalized functions into our control boxes, allowing users to experience differentiated functions.
Based on differentiated functional experiences, the future direction of automotive brands may split into two categories: one developing into automakers like Volkswagen and Toyota, and the other into automakers like BMW and Mercedes-Benz, targeting different audiences. The key is to find differentiated value points, especially experiential value, which is actually all about software experience differences. For example, if I want to give them a strong acceleration feel, I can set the software calibration data to be steeper. If I want it to drive more smoothly, I can calibrate the curve to be smoother.
Huan Shi: Data determines the ultimate product experience. We are already thinking about future user experiences.
Ding Wenchao: Yes, it is actually more about user thinking—how to combine user thinking with user usage scenarios to create some differentiation in software experience. I think this might be a future pursued by more automakers. Of course, for large-scale automakers, the main focus is on controlling software costs.
Huan Shi: So I feel that doing business in China is really not easy. Besides the problems everyone needs to think about, you also have to consider the customers of your customers. What kind of layout and planning do you have in terms of products and services?
Ding Wenchao: We mainly layout from the following aspects:
First, building highly reliable and cost-effective hardware platforms so that customers' software can run stably and support more market-oriented application scenarios.
Second, expanding the product matrix to cover various forms from distributed to domain control, making the variety richer.
Third, continuously upgrading the toolchain, evolving from the traditional V-model development to the software development model of the smart vehicle era, enabling better implementation of algorithms based on user demands, user scenarios, and data-driven approaches.
Fourth, strengthening service capability building. For small and medium-sized customers with limited software development capabilities, our team can provide professional support to help them complete software development.
Ding Wenchao: Finally, actively empowering industry chain talents. We cooperate with many universities to assist in talent cultivation. For example, we support over 100 racing teams such as the Formula Student teams, allowing students to exercise their software development and innovation capabilities in practice. Students are very passionate about competitions. We hope they can bring such passion to the entire automotive industry, continuously innovating in software, so the future will be better and better.
Huan Shi: I see. Mr. Ding has given us a very comprehensive explanation, covering everything from technology to application. I feel you are a long-termist. I also wish you to lead Huahai Technology further and further, and wish your products to serve more and more customers.
Ding Wenchao: Thank you.
Interview | Huan Shi Editor | Xiao Qie Review | Huan Shi