Author / Cao Jin
"Plans can't keep up with the changes." These simple 7 words have been vividly interpreted in all walks of life in the two years since the beginning of the epidemic, and the automotive industry is no exception.
With the postponement of the Beijing Auto Show and the control of the epidemic in Shanghai, the release of the Jidu concept car originally scheduled to be unveiled in April will also be postponed. However, Xia Yiping's mentality was still very stable.
Compared with the anxiety when Mobike launched the first batch of shared bicycles in 2015, Xia Yiping now senses more certainty about Jidu's first "car robot" product.
After returning to the automotive circle after 6 years, Xia Yiping re-internalized and sublimated his experience in automobile companies and Internet companies, transforming it into a wealth of a new competitive environment.
(This interview is online.)
Left: Xia Yiping, CEO of Jidu Automobile,
Right: Auto Byte head Cao Jin)
From the information revealed by Jidu Automobile, whether it is the development model of software and hardware decoupling, the smart capsule chip of Qualcomm 8295, or the SIMUCar (Software Integration Mule Car software integration simulation model car), it represents that automotive products are conditionally abandoning the tradition of "standardization" from demand and research and development.
Squeeze out "living time" on the "dead standard"
The online interview was scheduled for noon, because Xia Yiping did not have a lunch break - the epidemic disrupted the release node of core product information, and the pace of racing against the clock needed to be more intense.
"The recent epidemic in Shanghai has had a greater impact on the progress of the production of the exhibition car. For a better offline experience, we may delay the release time appropriately, but it will definitely be earlier than the Beijing Auto Show." Xia Yiping said.
In the past year, based on SIMACar's agile R&D model, Jidu completed the dual-domain integration of autonomous driving at high speed and in the city only 10 months after its establishment. Although progress has been very rapid, Xia Yiping said that Jidu has not blindly "sought speed".
"The reason why people feel fast is because they ignore a lot of the basic work that Jidu did before the brand debuted. In terms of vehicle development, we strictly follow the R&D rules and required cycles of automobiles; in terms of software research and development, in order to solve the problem that software research and development needed to be based on vehicle research and development in the past, resulting in a lag in software research and development, Jidu pioneered the development mode of SIMUCar software and hardware decoupling, and advanced the software research and development time of automatic driving and intelligent cockpit." Xia Yiping believes that the most core factor in greatly improving the efficiency of R&D is "talent".
"Whether it is the Geely team or the Baidu Apollo team, the synergy between them and the three parties of Jidu is very high, and everyone is highly consistent in their goal of creating a benchmark automotive robot product."
In addition, in the integration test of vehicle software and hardware, Jidu is also optimized based on the purpose of improving research and development efficiency - in terms of the traditional standardization process, car companies need to make the A-sample of vehicle hardware (prototype) first, and then do integrated software testing. From the very beginning, Jidu has carried out software and hardware decoupling research and development, so that every time the parts are prototyping, it can directly realize the verification of automatic driving and intelligent cockpit application software on the hardware.
"Taking software as an example, in order to obtain sufficient software development time, when the Qualcomm 8295 chip has not yet left the factory, we will first develop the technology on the 8155 chip. When the 8295 chip is in place, the previously tested software is ported to the new hardware, so that the overall R&D efficiency is greatly improved."
For the standardized R&D process used in the traditional automotive industry, Xia Yiping believes that "product quality standards are the lifeline of an automotive brand, and the quality standards of software and hardware must be strictly controlled, but the product development process can be optimized." He believes that through the decoupling of software and hardware, improving research and development efficiency, and promoting delivery nodes, Jidu can not only ensure the quality of the whole vehicle with strict standards, but also flexibly optimize the process in exchange for sufficient product development and verification time.
"Tradition" is the past tense,
But it's also the "wealth" of the future.
From Tesla's "self-research on everything" in the early days, to the current OEMs and suppliers have begun to embrace change, in Xia Yiping's view, it shows that everyone is jointly defining new standards to promote the evolution of the entire industry.
"In the past, parts production paid attention to standardization and scale, but once too much emphasis was placed on standardization, when new technologies and new trends appeared in the industry, the adaptive adjustment of products would not be agile enough, which may lead to a brand missing the opportunity for pioneering technology applications."
"Breaking the tradition" is more like a kind of thinking than a denial of traditional values. Xia Yiping's experience in ford and fiat has allowed him to gain valuable experience from the field of traditional car manufacturing.
In 2012, the Ford team led by Xia Yiping began to develop the Sync car engine system, which seemed to be ahead of its time. However, the process of software development can conflict with the inherent standards of electrical and electronic architecture.
"I did 6 years of internet of vehicle research and development, and the evolution of the field was relatively slow at that time." He said that according to the traditional process, a system usually takes three years from start-up design to delivery. However, during the past three years, technology trends are also constantly changing, so it is not a simple task to apply the latest and most cutting-edge technology to mass production vehicles.
Despite this, Xia Yiping still believes that the experience he has accumulated in traditional car companies is precious. "Experience is the mature cognition of things, and with the changes of the times and the innovation of technology, the experience accumulated in the past will be transformed into new wealth. For example, we still adhere to the principle of quality-oriented product design, inheriting the industry's standardization specifications in supply chain procurement, vehicle quality control and other aspects, which are the experience accumulated in the past experience."
He stressed that the traditional automotive industry is not "the past", and the experience of vehicle research and development and systematization that has been accumulated is extremely valuable. However, in order to find the optimal solution for the current business environment, technology environment and product evolution environment, it is necessary to reconstruct the past cognition.
How important is SOA?
In June last year, Jidu first proposed the self-developed electronic and electrical architecture for high-level automatic driving and the integrated JET 1.0 of SOA, that is, jidu intelligent architecture JET (JIDU Evolving Technology), which includes L4 level electronic and electrical architecture and vehicle-level operating system SOA, which add up to the nervous system of "automotive robot".
The birth of JET 1.0 stems from Jidu's consideration of the mass production of high-end autonomous intelligent vehicles. Xia Yiping believes that under the highly intelligent requirements of automobiles, how to deploy the core underlying hardware capabilities for products is a very critical issue. "For example, when we use an 8-megapixel camera, we need to match it with enough bandwidth to transmit video streams, otherwise it will inevitably affect the user experience."
The JET architecture for L4 autonomous driving has no experience to learn from in the industry, and its complexity is reflected in the need to connect all the domain controllers and sensors in the vehicle.
"SOA is an important technical foundation for decoupling hardware and software." Xia Yiping said that in the SOA environment, the electrical and electronic architecture is more likely to act as a bridge: "SOA enables a large number of instructions to be transmitted directly through the central brain to the edge actuator, so that the role of the electrical and electronic architecture will be transformed into a bridge between the hardware and the corresponding sensors."
At present, JiduJET1.0 has completed verification and realized hardware and software integration and adaptation on its software integration analog prototype SIMUCar 2.0.
"The decoupling of software and hardware development, the two two-line parallel research and development, such a change can ensure that each development link, at different stages of development, can be timely and fully verified." This is also a way to ensure that JET can effectively cooperate and adapt with software and domain controllers." Xia Yiping introduced.
Uncertainty and certainty of the environment
"This is the best of times, and it also needs to be looked at dialectically."
The automotive ecosystem is being reshaped. Smart cars face the best development opportunities, but at the same time, they must also face the problems that have never been encountered in the "golden age" of automobiles.
Since last year, there have been repeated news of suspension of production and price increases in the car circle.
In this regard, Xia Yiping is optimistic: "Due to the shortage of raw materials and chips in the automotive industry, the cost of automobile production and manufacturing in 2022 remains high, and it is expected that this situation will be alleviated in 2023."
Xia Yiping believes that the supply chain of smart auto parts follows the Matthew effect, and once the price of smart parts is reduced, a large number of companies will enter the field. The more players there are, the greater the purchase volume, and the price of components including lidar, cameras and chips will continue to fall. Therefore, by 2023, when The first mass-produced automotive robot is delivered, the supply chain environment it faces may have been improved.
In addition, the business model of smart cars will no longer be a single car sales. In Xia Yiping's view, the commercial value of software subscription services will go beyond the hardware itself.
"If you only treat the car as a single commodity, it can only maintain a gross profit of about 15%, or a net profit of 5%. When adding software services for smart cars, the business model will be completely different, and the user value will be redefined."
"Instead of doing L2++, we will apply L4 energy to power production"
"The product route chosen by Jidu is not to do hardware embedding, and it has achieved complete positive development in terms of system design, hardware selection, computing power support, etc." This is fundamentally different from the product route of first supporting hardware and then upgrading through OTA."
After undergoing the integrated development and operational verification of the software, Jidu's mass production models can bring users a high-level autonomous driving experience in multiple scenarios when they are delivered. "If it's just to build an L2++-level product, or assisted driving in a specific simple scene, there is no need to carry Orin's chip." We hope that the overall architecture of the product can meet and support the requirements for continuous evolution to high-end autonomous driving experiences."
The intelligent driving function that Jidu Automobile is about to land is simply "three domains in one" - high-level automatic driving that integrates parking domain, urban domain and high-speed domain.
Of course, if you think of Baidu Apollo's self-driving solution, you can't default to the technology of Jidu that is equivalent to Apollo's driverless. Xia Yiping said that Jidu Automobile is based on Apollo's autonomous driving capabilities, but it is more focused on providing users with better high-end autonomous driving mass production products.
"Jidu adopts an architecture based on Robotaxi's verified fully unmanned autonomous driving capability, which can be applied to high-end autonomous driving products that can be mass-produced."
He said that from the perspective of some autonomous driving perception decision-making algorithms and adaptation scenarios, Jidu's products fully apply Robotaxi's atomization capabilities, and what kind of experience and capabilities the product ultimately wants to provide determines the selection and architecture of its hardware.
"To arrive at the optimal solution of the system, three problems need to be solved - 1, consumer demand; 2, laws and regulations; 3, cost. This combination of optimal solutions requires us to constantly think and polish in the product design process, as well as to predict future industry trends based on our understanding of laws and regulations."
Recently, European and American countries have gradually relaxed restrictions on automatic driving, which has made domestic practitioners also begin to think about how to more forward-looking layout of new technology research and development, to create a high-growth product system.
Xia Yiping added that Jidu is not directly to bring users a fully autonomous driving product, because the current laws and regulations are not yet allowed, but also fully combined with the actual needs of users. "We need to find the best balance between multiple factors."
The real intelligent competition has not yet opened
"In the past five years, the automotive industry has completed the evolution of vehicle electrification." For the current new generation of car manufacturing, including Jidu, Xia Yiping believes that the "post-electric era" needs to shoulder new missions and accept new challenges - to promote the process of automobile intelligence and open the era of intelligent vehicle 3.0.
"The early brands have promoted the transformation of the entire industry to the direction of electrification, triggering a global Matthew effect - whether it is the government's policy inclination, or the increasing acceptance of new energy vehicles by consumers, as well as the maturity of the supply chain and industrial chain behind it, the electrification of automobiles is destined to become a historic innovation in the automotive industry."
In this regard, Xia Yiping said for example, 3 or 4 years ago, the mainstream brands in the market could still take the acceleration performance and mileage of electric vehicles as the core competitiveness of the product, but in just a few years, the industrial pattern has undergone earth-shaking changes, and this change is due to the maturity of the upstream and downstream supply chains of the entire industrial ecology. "The vigorous development of the industry is inseparable from the development of the industrial chain and supply chain, and scale and standardization are the core essence, but from another point of view, the standardization of the supply chain will also erase the barriers to competition."
In his view, the first year of automobile intelligence is 2023. The ultimate contest of intelligence is actually "unmanned" automatic driving. When unmannedness comes, we can truly liberate productivity and the car can truly transform from a tool attribute to an intelligent mobile space attribute.
※"Chief Chi Heng" is a high-end interview column launched by Auto Byte. This column tells about inviting representative figures in the field of intelligent travel, expressing views in depth and recording the stories of the times.
About Auto Byte
Auto Byte launches the Automotive Technology Vertical Media for The Heart of Machines, focusing on cutting-edge research and technology applications in the fields of autonomous driving, new energy, chips, software, automobile manufacturing and intelligent transportation, and helping professional practitioners and relevant users in the automotive field understand technology development and industry trends through technology to gain insight into products, companies and industries.
Welcome to follow the stars and click Likes and Likes and Are Watching in the bottom right corner.