Recently, ZhenFund held a closed-door event called "ZhenCraft·Intelligent Automobile Roundtable" and invited colleagues from intelligent driving and automobile industries to discuss the opportunities and challenges of the automobile industry chain under the wave of intelligence, from the perspectives of experts in various industries such as automatic driving chips, OEMs, sensors, new forces, multinational component leaders, and intelligent cockpits.
"In this event, we have heard many different discussions among practitioners, but the same belief is that technological progress will create new value. We will always be the co-pilot, accompanying and supporting entrepreneurs in their journey to change the world."
-Yao Fangzhou, Managing Director of ZhenFund.
"The development of technology innovation to industry landing is not an overnight success, and progress always accompanies stage questioning and hesitation. ZhenFund has always believed that the intelligentization of the automobile industry will make a huge contribution to human civilization in the long term. We will also continue to support innovators in the automobile industry, bringing solid products and services to the industry on the innovation road with technology as the main line."
-Zhong Tianjie, Investment Director of ZhenFund.
Skateboard Platform——Create an integrated software and hardware automotive operating system
Sharer: Cai Dexuan, Vice President of U POWER
Skateboard platform helps build cars freely
When talking about skateboard platform, we must first talk about cars. The functions of a car can be divided into: walking, carrying, and safety. The skateboard platform is a complete lower body that can independently realize the walking function. In the development of upper and lower split cars, the skateboard platform bears 80% of the safety responsibility, which is a technical difficulty and a key point to making the upper and lower parts separate.
We believe that the skateboard platform has three characteristics. First of all, it must play the role of physically supporting the upper vehicle body; secondly, it must be able to keep up with the vehicle body to perform physical and information links very conveniently and quickly on the premise of independently realizing the running function; finally, it needs to undertake the whole responsibility for the safety of vehicles and property.
U POWER focuses on creating a standardized skateboard platform, so as to empower the scene owner to define the car, promote the diversification of ecology, and finally achieve the personalization of user experience.
UP Super Platform: Use the ultimate standardization to achieve the ultimate personalization
At present, most domestic autonomous driving companies use Lincoln MKZ as a prototype, but it is difficult to switch to other models, because this requires autonomous driving companies to deal with all actuators, steering, braking, etc., and they need to do it themselves. Controlling these things is very troublesome. The software-hardware integrated automotive operating system for the era of autonomous driving is like human limbs, torso, cerebellum and brainstem, helping us solve this problem. This is actually a key function that a functional car must have to evolve into a smart car.
U POWER's vehicle operating system integrates functions with high real-time and reliability requirements such as motion control, thermal management, three-electric system, and sensor input, which is equivalent to the "little brain" of a car. This not only provides a common platform for the "brain" of the car, but also allows autonomous driving companies to easily switch from MKZ to other smarter vehicles, enabling OEMs, smart driving developers, and operators to focus on algorithms, personality areas that focus on user value, such as globalized application development and user operations.
In addition, the standardized skateboard platform can realize the upper and lower split development of the car, so that the upper and lower car bodies can be developed separately and decoupled, and can be carried together as a whole. Through this model, the entire development cycle can be shortened by 6 to 12 months compared to before. Based on a standard skateboard chassis, the entire process from vehicle development to mass production, which previously took at least 24 months, can be completed within one year.
The ecological cooperation of "hardware + software" empowers the industry
U POWER insists on making cars for the scene, formulates the hardware structure and parts standards of the skateboard platform, and provides the whole vehicle operating system. Not only can it integrate the supply chain for OEMs and provide pre-integration of components, but it can also provide auto-driving developers with a vehicle-specific computing power platform and standardized motion control, and it can also serve as a sales channel with battery suppliers. The ecological cooperation of "hardware + software" can save developers development time and reduce development costs, thereby forming a large-scale and continuously empowering the automotive industry.
Data closed-loop defined chip
Sharer: Zhang Jianyong, co-founder of Rhino
Establishing an intelligent driving system is essentially establishing a self-enhancing data closed loop
As a highly computing-intensive application scenario, intelligent driving puts forward higher requirements for computing power and data processing capabilities. Against the backdrop of the gradual failure of Moore's Law, how can chip companies respond to the challenge of intelligent computing and support the evolution of intelligent driving to a higher level? This requires overall planning and systematic advancement.
The transformation of the automotive industry from software empowerment to software definition has increased the importance of full-stack closed-loop capabilities. If artificial intelligence is compared to a rocket, the data is the fuel, the algorithm is the engine, and the computing power determines the upper limit of the engine's capability; Greater computing power, better performance of the entire system, promote mass production and deployment of more vehicles and encourage users to use it to generate more data. In essence, it is to establish a complete closed loop of computing power, algorithms, and data to form an intelligent iterative system that continues to grow and gain itself. Whoever iterates faster will have a competitive advantage, and whoever turns the closed loop first will gain greater momentum. This has been successfully proven in the fields of Internet search recommendation and video analysis and has generated huge value.
The underlying capabilities of intelligent driving are essentially perception, cognition, and decision-making capabilities. To do a good job of these underlying capabilities is not just to make a better chip, but to define it from the new paradigm of intelligent computing, which is data closed-loop. Rhino created the "data closed-loop definition chip", with the core of building a self-enhancing closed-loop of computing power, algorithm, and data, focusing on the final product architecture and functional design, creating a more powerful computing power driven, larger data scale, an intelligent system for faster algorithm iterations.
Empower smart travel with an innovative computing platform
Throughout the domestic OEM autopilot market and technology trends, domestic OEMs are very urgent to upgrade autopilot functions: from basic L2 functions to high-speed NOA, and further challenging urban NOA, the AI algorithm capabilities, system architecture capabilities and Engineering mass production capabilities pose a huge challenge. At the level of autonomous driving hardware, the introduction of multi-source heterogeneous sensors, especially the rapid upgrade of camera performance, generates massive data and requires a sharp increase in computing power; BEV large models as a new paradigm for the next generation of high-level autonomous driving perception algorithms have gradually become the consensus of the industry . The data-driven model of "big model + big data" has become the key to the rapid iterative evolution of autonomous driving technology.
Rhino is committed to empowering smart travel with innovative computing platforms , using high-performance autonomous driving chips as the base to drive larger data scale and faster algorithm iterations, and working with partners to create mass production solutions for advanced intelligent driving systems, helping car companies build chips The data closed loop of architecture and advanced algorithms makes machines smarter and smarter, and keeps autonomous driving safe, reliable and stable.
Panel
What changes will AI bring to the autonomous driving industry?
There are currently two mainstream views in the market: one is to use a single camera to implement basic ADAS functions to solve stability and cost issues; the other is to evolve from basic ADAS to high-end AD functions, and AD performance is improved through hardware pre- Buried and data closed-loop way to update and iterate. At present, the performance of AD has generally not reached an amazing level, and is still in the process of quantitative change. In the medium and long term, with the rapid iteration of closed-loop data, the improvement of autonomous driving performance and experience will shift from quantitative to qualitative . For example, Tesla’s FSD subscriptions in the United States have begun to rise. We believe that domestic leading companies will also catch up quickly is the general trend of the industry.
——Zhang Jianyong, co-founder of Rhino
Despite the recent attention on generative AI models such as ChatGPT, they may not be able to solve some key problems in the autonomous driving industry. Autonomous driving incorporates perception, reasoning, and decision-making technologies that require high levels of precision and explainability to enable more sensitive and less error-prone decisions than human drivers. At the same time, an automated driving system needs to be aware of its functional boundaries so that a human driver can easily take over.
——Wen Licheng, young researcher at Pujiang Laboratory
The introduction of AI technology into autonomous driving is a major trend, and it may take a year or two to establish itself. On the vehicle system, the combination with the vehicle-machine system may be the first application point . In addition, the application of this technology in controllers is also possible , but it is questionable whether it can improve reasoning or structuring capabilities.
——Zhang Yixin, machine learning development expert at Google
For your own segmented field, what do you think about the domestic substitution of core technologies?
In the EDA industry, we should not only be satisfied with domestic substitution, but should learn to imitate and accumulate experience to surpass latecomers and strive to achieve global leadership. At the same time, in some cases, technology can be exported overseas and realize independent counter-control. Because in the EDA industry, the technical threshold is relatively high, which requires long-term accumulation and research and development, and we cannot only rely on domestic substitution to solve the problem.
——Xu Yongxin, COO of Chip Easy
I think "industrial opportunities" are more important than "technical corner overtaking" , especially in the EDA industry that needs to consolidate basic science. Find opportunities for industrial transformation, follow industrial transformation and upgrading with the spirit of craftsmanship, and continuously iteratively upgrade technical tools to polish good EDA products.
——Huang Wu, Senior Product Director of X-Epic