*Author: Chris Zheng
On September 15th, ideal Automobile announced that Wang Kai has been appointed as its Chief Technology Officer (CTO), responsible for the research and development as well as mass production of intelligent-related technologies of ideal Automobile such as electronic-electrical architecture, intelligent cockpit and autonomous driving.
Internally, Wang Kai holds a high rank in the management hierarchy of ideal, unquestionably being the most important executive recruited by ideal over the past five years.
The outside world is full of curiosity about this personnel appointment and even about Wang Kai himself: why Wang Kai for ideal? Why ideal for Wang Kai? And does an automotive enterprise need a CTO position? Let’s explore them step by step.
Why Wang Kai?
In the two-hour communication, Wang Kai mentioned “landing” 20 times and “Tesla” 18 times. Tesla is just the company with the most experienced in landing intelligent automobile technology.
In his past career, Wang Kai’s expertise lies not only in theoretical research but also in the long-term accumulation of landing experience. This is probably one of the most valued character traits of Wang Kai by ideal Automobile.
In the interview, Wang Kai stated that a car company truly transformed into a technology company is marked by “having a particularly strong integration capability for various subdivision technologies and basic sciences and being able to land them truly. This is the highest condition.”
Applied to the trend of centralized automobile electronic-electrical architecture, this means that car companies need a deep understanding and recombining of different resources such as chips, algorithms, operating systems, and hardware structures throughout the vehicle.
In 2012, Wang Kai jumped from Nokia to NXP Semiconductors as a senior hardware engineer. A year later, NXP Semiconductors initiated a SmartCore™ cockpit domain controller development project with Wang Kai as Chief Architect. By the end of 2017, the project had landed successfully in the new Mercedes-Benz A-Class (MBUX Intelligent Cockpit).
This is the world’s first mass production project for a cockpit domain controller. After that, Wang Kai led five different mass production projects in various directions, successfully expanding his career into different aspects of automobile intelligence such as chips, algorithms, operating systems, and hardware architecture.
This gives him better control over the technical details of the underlying technology. For example, except for Tesla, Ideal is the second car company that explicitly states its intention to independently develop real-time operating systems for autonomous driving.
When explaining the logic of self-developed operating systems, Wang Kai mentioned a core indicator of autonomous driving systems- latency. Ideal must independently develop the kernel part of the real-time operating system and system middleware to ensure the optimization of the entire chain in the future.
To ensure the ultimate user experience- more accurately, when users provide feedback, diagnose problems quickly from the source code level, and update quickly with OTA-, Ideal needs to develop and control the entire technology stack.In addition to his expertise in implementation, another characteristic of Wang Kai in the field of landing is his sensitivity and judgment towards macro technology trends.
At the beginning of the conversation, Wang Kai mentioned the bleak prospect of 14 first-tier suppliers in the automotive industry, including Visteon where Wang Kai was located, suffering losses. However, on the other hand, automotive technology companies (referring to emerging data-driven car companies such as Tesla, LI, NIO, and XPeng) have been recognized by both users and Wall Street in terms of differentiation.
Wang Kai believes that the differentiation of these “automotive technology companies” is by no means limited to more investment in the cockpit and assisted driving systems. From the perspective of basic science, the wave of change in the automotive industry represents the intersection of multiple basic disciplines, which has promoted the industry’s new round of revolution.
Communication: The evolution from 4G to 5G, latency reduction, cost reduction, and throughput increase.
Industry 4.0: Manufacturing automation, cost reduction, and efficiency improvement.
Artificial Intelligence: Empowering the cockpit and autonomous driving.
Big data: User data-driven product iteration.
Material science: Automotive three electro technology and lightweight.
Wang Kai incorporated his macro judgments on the development of different industries into his career planning. In June 2012, the pioneer of intelligent cars, Tesla Model S, debuted in California. Wang Kai sensed the hidden trend of the automotive industry, and five months later, he joined Visteon to lead the implementation of the new generation of automotive electronic and electrical architecture integration.
The combination of the double characteristics made Li Xiang say in the announcement: Wang Kai is the most suitable and competent candidate among many candidates as the CTO of ideal car.
Why is it ideal?
Before answering this question, we have to briefly introduce Wang Kai’s background. Prior to joining Ideal, Wang Kai had worked overseas for 15 years.
So, this question may need to be divided into two sub-questions: why did he return to China? Why is it ideal?
Wang Kai talked about the market pattern in the smartphone industry where one dominates others: Apple is the only disruptive innovator among the major companies, and all the other leading companies, except Samsung, which has the advantages of the entire industry chain and intensive effect, are Chinese companies. Huawei, Xiaomi, OPPO, and vivo all quickly followed in the footsteps of Apple after it ignited the smartphone market.
Why are all Chinese companies? Because China has three characteristics of the intelligent device market: a large market, many users, and data-driven. These three characteristics also apply to the smart car market.
So the problem became clear: the disruptive innovator in the smart car industry is already there, and what Wang Kai is looking for is a Chinese fast learner who shares his values.
“I have always supported Progressive innovation, which is a step-by-step technological revolution. By solidly taking each step and implementing the technology, and then replenishing the next step of research and development after the technology has been implemented, we can gradually move forward,” said Wang Kai.# The Ideal ONE car and the role of CTO
The Ideal ONE car is filled with the shadow of a “progressive technological revolution”. For example, instead of relying on expensive and inefficient electric powertrains like those found in most electric vehicles, the Ideal ONE car combines electric and fuel power through the use of a less mainstream but practical powertrain design to achieve both smooth electric driving and convenient fuel supplementation.
In the eyes of Wang Kai, this is the most suitable product decision from the user’s perspective.
Wang Kai also fully agrees with Li Xiang’s proposals that “specific actions should be taken at specific times” and “the strategies of a company from 0 to 1, from 1 to 10, and from 10 to 100 are entirely different.”
“Many companies strive recklessly, which presents a significant risk.” Ideal sets specific goals for each stage and only moves on to the next phase once the previous phase is solidly established.
To a certain extent, this “stage theory” has already been reflected in the Ideal Cars autonomous driving department. With the launch of the L4 autonomous driving car project in 2020, the autonomous driving department of Ideal Cars has begun to expand steadily. According to Wang Kai, the autonomous driving department’s team will increase from 60 to about 200 people.
“Being precisely wrong is far deadlier than being roughly right.” Wang Kai remarked.
Does a car company need a CTO?
“I used to be something of a ‘golden player,’ but at Ideal Cars, I have to be a ‘star coach,’ leading every member of the team to become a ‘star player,'” said Wang Kai, explaining his perception of the CTO position.
Ideal Cars wants to become a data-driven tech company. Wang Kai is responsible for designing the overall architecture, which is no longer just the narrow automotive electronics and electrical architecture or the overall vehicle architecture, but an enterprise-level closed-loop architecture.
From 0 to 1, Ideal Cars completed the closed loop of the entire vehicle, mobile app, after-sales system, R&D system, and cloud. When user needs are adopted, feedback can be quickly provided to R&D with real-time progress, estimated release time, and more.
Wang Kai’s role in taking Ideal Cars from 1 to 10 is to push the entire closed-loop into a flywheel iteration, making the entire iteration faster and faster.
Wang Kai said that to achieve a flywheel effect and accelerate growth in all departments, all small flywheels must combine into a macro flywheel that can sustain sales growth, mass user feedback data, continual R&D improvements in user experience, and ongoing sales growth.
In order to facilitate cross-department coordination throughout the process, Wang Kai oversaw the intelligent cabin, autonomous driving, and newly added computing platform departments.
From the perspective of automotive electronic and electrical architecture, there are three very important trends in the industry: centralized computing, satellite data gateway, and satellite energy gateway.During this evolution, the Ideal Automotive Computing Platform department has increased the capacity of the self-developed ECU, promoted the integration of data gateway and energy gateway, and prepared for the abstraction of the entire hardware for the next stage.
On an enterprise level, Ideal Automotive will ultimately become a travel service provider during this process. From Wang Kai’s point of view, such enterprises are absolutely technology-oriented.
“Does a technology company need a CTO? You think about it.”
In fact, there is nothing that can better illustrate that Ideal has entered the stage from 1 to 10 than Wang Kai’s joining. We also look forward to Wang Kai’s flywheel effect, which can bring about a new Ideal through fission evolution.
Media: How many people do you plan to build the R&D team in the next stage?
Wang Kai: Specific actions should be taken at specific stages. When going from 0 to 1, it’s more about how to get the mechanism of the matter started. It’s necessary to control the cost. It’s a huge waste to invest money without figuring out the direction.
In the stage from 0 to 1, Ideal Automotive is focused on operations, not reducing the salaries of R&D engineers. From stage 1 to 10, I can say responsibly that we will definitely have higher salaries for R&D personnel than the market, and some individuals will be offered much higher salaries to join us.
Media: What is the scale of R&D investment per year?
Wang Kai: Let me talk about the number of employees first, because the money depends on what we are going to do, and the actual employees will have a big impact.
For example, the autonomous driving team, because we need to develop it ourselves, used to use Mobileye stuff, and perception was not self-developed. Now we want to do perception ourselves. The autonomous driving team will expand from a small scale of 60 people to 200 people.
Media: How many people do you have now?
Wang Kai: 60. This is just one area. From a long-term perspective, we have a lot of layouts. As an OEM, we need to reach far with the tentacles of technology companies. We value the quality of talents. You can see that Tesla is the same, not just a bunch of people.
There are opinions that it takes tens of thousands of people and billions of lines of code to do this kind of software. We have a different view on this point. We think that people at Tesla are very sophisticated and of high quality.
From my perspective, efficiency has always been the focus. Efficiency is the most core aspect. Your efficiency in the use of capital, personnel, and operations is the most critical. We hope to have high-quality teams composed of top talents in various fields.
Media: How do you define core technology?
Wang Kai: Core technology is something that can directly reflect value, firstly, it is data. As I just said, data-driven technology companies consider data as the gatekeeper. The other is related to users because it is to improve the ultimate experience of end users.Using autonomous driving as an example, one of the biggest challenges in the field is latency, which is a very significant one. In order to control this, you need to have control over the entire technology stack. The core kernel component must be in your own hands to ensure that latency is controllable.
Once you hand it over to your supplier, their development becomes a black box, and you are cut off. When a customer requests a change, you are unable to fulfill the request.
While the OTA mechanism itself has been established, if something is a complete black box and cannot be modified, or if you need to pay to make changes, even if you pay, you cannot quickly iterate, and then you are at the mercy of the vendor. For us, the architecture of full-stack software needs to be controlled by ourselves.
TheMiddleware’s kernel also needs to be in our own hands, and we will build APIs for these two kernel parts, especially OEM’s kernel where we will introduce reasonable APIs for third-party participation, and some ready-made things can be incorporated.
Through the external API built by the Middleware’s Kernel, ready-made algorithms can be introduced. This set of things becomes a system, with its own Middleware, and coupled with the upper layer application, it can be used directly through the API. However, the most important data interaction is organized by us.
For our partners, we hope to deliver a white box so that we can truly control the feedback from the client. We can quickly diagnose where the problem is, update it, and complete the loop with the user.
Media: What is a white box?
Wang Kai: Source code.
Media: What is the difference between the core and the whole kernel of OS and Middleware?
Wang Kai: The kernel is essentially a cornerstone. The kernel part includes file systems, IO systems, and also boot, which is the very beginning of booting. This is the most critical part. For example, some applications and even drivers that extend from the OS range do not belong to the kernel part.
Media: Are they not that important?
Wang Kai: They are equally important, but this part does not necessarily have to be self-developed for everything. I can even hand it over to a third party because it is not the lifeblood. If you follow my rules, I can guarantee myself.
In the automotive industry, especially in autonomous driving, the most important thing is the worst case. Tesla does a lot of things quickly because everything is self-developed, including self-developed hardware, because it feels that no one can meet its software needs.When software engineers reach the pinnacle of their career, they will definitely want to work on hardware as they are not satisfied with what is currently available. Similarly, top hardware engineers would also want to work on software and claim that others are too low. This is a natural progression.
Media: Previously, Ideal had a roadmap for autonomous driving. Since you joined, have there been any updates to the roadmap or timeline?
Wang Kai: The process remains the same. First, we will implement NOA. I don’t want to talk about so-called L2 and L4.
The user’s experience depends on the scenarios. With more scenarios, the autonomous driving time and the availability of assistance or autonomous driving become more critical factors. We hope to improve this, including implementing data feedback loop based on user demands, and downplaying the importance of L2 and L4.
Regarding scenarios, some have described it dogmatically, such as Tesla’s NoA, followed by L3, L3plus, with increasingly broad application scenarios before reaching L4.
Important milestones include cars released in 2022, which require ready L4 hardware. It is not about installing all hardware at once, but rather ensuring that all interfaces are in place. The car can be upgraded not only through software updates but also hardware upgrades.
Media: Does Ideal have the qualifications for surveying and mapping? Do you plan to achieve high-precision maps in the future, which may require the use of supercomputers?
Wang Kai: Yes, we have the qualifications, level B qualifications (for internal use only, cannot be sold externally).
From my perspective, data is the most important aspect. I have always said that the company must be driven by data. Maps, including high-precision maps, are essential data, although it is not user data. When your technology is strong, the cost of making cars is minor compared to other areas. Traditional technologies, such as mechanical manufacturing and engine technology, are more mature. However, the integration of these mature technologies is crucial and requires considerable investment. This is how your competitive advantage can be demonstrated, and the user can see your iteration speed.
Tesla’s supercomputer is at the forefront of this field. However, we also have our advantages. In my opinion, if Tesla does not build a robust local team in China, it will most likely be defeated by local companies.
A critical factor I mentioned earlier is that the future of the automotive industry is driven by data-driven technology companies that must have strong local user connections, especially in the field of autonomous driving. Many data points have different regulations across countries, which can be difficult to understand when you are from outside of that country. Local companies are rooted in their home country, and their employees are born and raised there. Therefore, they understand what regulations they must follow, and their iteration speed is inherently faster than outsiders.Tesla brought many new ideas, and brought the entire supply chain into play, which indeed has a pioneering role.
Media: How do the cabin, autopilot, and computing teams work together?
Wang Kai: From the perspective of the automotive industry, there are three particularly important trends in the concept of architecture. The first trend is centralized computing, the second trend is satellite data gateway and energy gateway. Don’t know if you have heard of the Zone Controller? It is the implementation of the three trends I mentioned earlier.
Let me briefly explain, and then answer why the organizational structure should be matched, which is relevant.
Centralized computing is very simple. In the past, the manufacturing industry was a function-driven industry. When I needed a function, I added an ECU, and the final result data, whether it was input or output, was transmitted to other places to complete this function. In the past, when I added a panoramic view, I just added an ECU, and it only provided you with a panoramic view, and that was it.
But what does today’s smart car require? When my product was defined a few years ago, the architecture was defined, and the initial version and basic functions were provided in the form of software. In the past, after a few years, the car would be like this. But now, after I give it to you, your software part is constantly being upgraded and becoming better, and it will mobilize all internal resources. At this time, the original function-driven architecture will no longer work.
For example, the panoramic view function we just talked about, any intermediate result, because this is done by the supplier, it’s just that I give you power and data cables, you output it, and that’s it. Any intermediate result cannot be used.
If you want to use it, I have to go to find them, and you have to make a small change, and the supplier says it’s okay, one million dollars, three months, then how can you do OTA (over-the-air upgrade)?
So the best solution at this time is to use hardware to build a big framework at the beginning, and once the software’s foundation is complete, try to calculate things to the core. And these sensors are just data coming in.
As for how I use it, there is a pairing combination given to the internal parts on the central computing platform at the beginning, and when there are new ways to play after the upgrade, just change it in the middle. Everything on the end doesn’t need to be touched, and this is completely different from the previous concept.
Earlier I mentioned two things: satellite data gateway and energy gateway. What do these two things mean?
In my original architecture, it was normal for a luxury car to have 200 ECUs, with many wires entangled, and it was very difficult to assemble. This no longer meets future needs, and you are completely updating, each with only the final result, and cannot change in the middle, with very limited functions. I have to beg and may not be able to achieve it. Once I solve one ECU, others can use it the same way.In the new experience, try to centralize all calculations to one or a few core ECUs with actual brains. Previously, many ECUs were distributed throughout the vehicle, and the things inside the vehicle were either inputs or outputs. Inputs were data gathered by sensors, while outputs were controlling a motor or opening/closing a window. These ECUs were placed in these locations because they were close to the motor. In the new generation, these will all be centralized into a satellite network, similar to a central city and satellite cities.
For example, Beijing and Tongzhou, what is Tongzhou doing? These people actually work in Beijing, but they live here. At the beginning, they rest here, and both the energy and data are collected here, and work is done through intelligence.
The benefits of doing this are manifold. Firstly, it solves the problem of easy OTA upgrades and management.
Secondly, it considers the physical layout, greatly shortening the layout route. Tesla said they wanted to shrink it down to 100 meters, but they haven’t done it yet. In this respect, Tesla is still relatively advanced.
Last year’s core patents, I hope to see them reflected on the Model Y, but it hasn’t been done yet. There are several key elements, such as 2D layout. Once the backbone is completed, everything is on it. The second is that all basic parts have communication interfaces, which are basically doors that are installed and plugged in.
Like these patents, two were issued last year, but they were not ultimately implemented on the Model Y. But actually, everyone recognized this concept and it is an inevitable trend in the future. The reason why I mentioned this is to return to why we organize our architecture this way.
Organization architecture is still the same, and the operation you do is also a technological iteration, step by step. Currently, there is an intelligent and system department, which mainly deals with everything related to intelligent cockpits, as well as services for cloud and mobile phones, including OTA channels.
There is a team dedicated to autonomous driving. The third is the computing platform, which will develop some core and particularly significant local ECU increases in capacity. We need to complete the fusion of the energy and data gateways.
Media: Self-developed ECU?
Wang Kai: Yes, the fusion of energy and data gateways. The next step for the platform department is to assume the core of entire hardware abstraction. At that time, the hardware will be completely abstracted. This is the next stage, step by step. In the future, various functional domains will be refined into requirements for the platform, with emphasis on software, and connecting all these things is a matter of systems architecture.The energy gateway is power supply. Before electric cars, there were many energy supply or fuel supply systems, which were very important for vehicle architecture. However, in the case of electric cars, this is the most important thing. How the electricity is distributed and its efficiency of use.
In fact, one of the reasons for Tesla’s in-house hardware development is that it believes other chips are not energy-efficient enough, which is a typical example of a very first-principles approach.
Media: How far is the gap between the ideal and the Model Y, and how long before it can catch up?
Wang Kai: Tesla’s central domain controller is basically there, and the follow-up is performance improvement. I am looking forward to what Hardware 4.0 will look like when it comes out. At that time, automatic driving and the cockpit will be truly integrated, in a chip, which is very powerful.
Currently, these two parts are separate, two ECUs, which is a necessary process. If you don’t do this part solidly, you will encounter obstacles at each step when defining product requirements.
Our next step is to make the interaction better, and our focus is on software and internal data closed loop at this stage.
In fact, software is the most difficult part of doing pre-controllers, especially the integration part. Function safety ranges from QM level to ASIL-D level, with a large range, and the complexity of the system is exponentially increasing. If your software is not solid, it is better to do this separately. Because no matter the complexity or the cost of research and development investment, enterprises cannot afford it.
When you thoroughly grasp both sides, and even make software and hardware separation during development, but the efficiency is integrated, then you can move on to the next step after this is completed, from 0 to 1 in your system, and enter the cost reduction phase.
If you haven’t completed the 0 to 1 step yet, doing the next step now will be fatal, because for automakers, many indicators need to be met, and we have clear time points for when this car will be launched, which are hard indicators.
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.