At present, the world's leading car companies have launched the development of a new generation of domain centralized electronic and electrical architecture, and actively deployed corresponding software capabilities.
For example, GeEP 4.0, a new generation of architecture that Great Wall Motor plans to launch in 2022, will evolve into three computing platforms: central computing, intelligent cockpit, and advanced autonomous driving, and adapt to a centralized standardized software platform based on SOA.
According to the monitoring data of the Gaogong Intelligent Automobile Research Institute, in 2021, L2+ began to gradually replace L2, and L3 began to enter a period of growth climbing. As the intelligent hub of the car, the advanced degree of the electronic and electrical architecture will directly determine the degree of intelligence and the space for evolution of the car.
"Gaogong Intelligent Car" learned that in addition to international Tier1 giants such as Bosch, Continental, ZF, Aptiv, etc., a large number of Chinese players such as Desay SV and Neusoft Ruichi have launched domain controller products for different levels of automatic driving.
However, according to the four stages of distribution, intra-domain fusion, cross-domain fusion, and central computing that the electrical and electronic architecture has gone through, most manufacturers are currently in the initial stage of intra-domain fusion and have begun to centralize decentralized data computing.
Today, "empowering" automakers as a supporter behind their customers has become a strategic transformation focus for traditional software vendors.
Looking at the entire industry, it can be said that no supplier can provide a complete domain control solution to support complex software and hardware development. In addition to the self-research of the main engine factory, most players basically take the core supplier as the center, jointly integrate the ecological alliance circle, and form a professional subdivision field.
Among them, as an end-to-end Tier1 player, you need to find many sub-suppliers to integrate a core technology, but this traditional Tier1 project operation method is a very big challenge for the coordination and management of resources.
The feedback from core technology suppliers is slow, and it is difficult to mobilize effective resources to respond in the first time, which are all headaches for OEMs.
In addition, in the face of the chip shortage during the epidemic in the past two years, such manufacturers are basically powerless, and their competitiveness has become a castle in the air.
Under the trend of software-defined automotive SOA, automotive software will go through two important stages: "soft and hard decoupling" and "soft and soft separation". The disruptive innovation opportunities of E/EA architecture change and the construction of a new rapid development and R&D model for SOA are the key factors in the digitalization and intelligent upgrading of OEMs.
Among them, suppliers such as Yingchi Technology, TTTech, KPIT and other suppliers that have the ability to provide third-party software platforms have become the mainstream choice for the automotive industry to create a next-generation vehicle electronic architecture.
"We are committed to empowering OEMs to lead in the process of intelligent reform. At the same time, after the partners of the parts get our products, they can greatly reduce the difficulty of matching the business of the car manufacturer, shorten their development cycle, and reduce costs. That's why we enrich the middleware and transform it into a software platform. Wu Junjie said.
For car companies, it is not easy for the existing limited software and hardware teams to dominate the penetration and collaboration between AutoSAR Adaptive and security middleware.
At the same time, for the development of new cars, the time is tight, the task is heavy, and the integration difficulty can be imagined. In addition, from single domain to cross domain (body domain, cockpit domain, autonomous driving domain), the complexity has increased exponentially.
For automakers, in addition to the underlying real-time operating system, AutoSAR, secure real-time communication modules, and other key software modules are still very complex, and in the case of the whole vehicle, various software running on different hardware platforms must work together.
The rapid growth of software complexity has greatly widened the gap between the industry's need to develop complex software and the automakers' self-research capabilities, which involves a large number of software function service capabilities and integration with secure communication modules. This gap can only be quickly closed in the short term through software reuse and adoption of third-party underlying software platforms.
The software platform EMOS launched by Yingchi Technology integrates the enhanced AutoSAR AP (adding self-developed deterministic scheduling and communication) and the integration of traditional CP, covering the vehicle central computing unit, automatic driving domain control, cockpit domain control (key functional safety part), and the entire architecture is defined for SOA.
At the same time, EMOS connects MCU, SoC and other chips and middleware, which can realize real-time, diagnosis and monitoring of all applications including scheduling and communication. In general, the platform is in the form of a service, which includes services for sensors, vehicle control, commissioning, etc.
With the upgrading of the function of intelligent driving, the domain controller will access more numbers and types of sensors, with the ability of multi-sensor fusion, positioning, path planning, decision control, communication, etc., and the real-time, reliability, and security requirements for massive data processing, communication transmission and vehicle control are very high.
On the one hand, in the process of multi-sensor fusion, the increasing number of sensors will bring about an explosive growth of data with different structures, and the calculation and fusion analysis of these massive data will test the software capabilities of manufacturers.
Different types of sensors have different algorithms, chip platforms, and different structures, which are placed in a "box", and the complexity of its internal structure can be imagined.
Therefore, how to properly layout and platformize the domain controller architecture is of great importance to domain controller vendors.
"So far, you'll find that very few OEMs in the industry are mentioning L3, because there is no mature technical solution in this time window to support the advent of L3 level semi-autonomous driving. Although in the past, everyone would have claimed that two controllers, ARM and the security cerebellum, could be used to implement L3, but it is clear that such a cost is difficult to be accepted by the main engine factory. Wu Junjie said.
For suppliers, cross-domain integration is still a difficult thing. This requires a complete and systematic understanding of the whole vehicle, and at the same time to maintain a close cooperative relationship with the main engine factory and chip factory, and has very high requirements for the quantity and quality of resources.
Positioned as a bridge connecting chips and autonomous driving technology, Yingchi Technology has been sparing no effort to establish a deep binding relationship with major manufacturers, and timely integrate the customer needs of automakers, and carry out cutting-edge research and development with many chip companies.
In June this year, the company announced the completion of nearly 100 million yuan of A round of financing, led by Sequoia China and SAIC Hengxu, and followed by Horizon Strategy. In October, we formed a global gold partner with NXP.
Its new DCU3.0 this year is equipped with the hot Horizon Journey 3 chip (responsible for the all-in-one solution) and the NXP S32G chip (which can cross the intelligent driving domain and the intelligent gateway domain at the same time), providing the maximum possible functional safety design, while enabling greater flexibility and scalability for different high and low configurations of the model.
"Intelligent driving perception relies on AI computing power, but at present, there is no kind of AI computing power in the industry that can reach the level of functional safety ASILD, so we will use different kinds of sensors from the system level to do perception fusion, and only when the results of several perceptions reach consistency, the system will recognize it." This is actually a more costly approach at the system level to achieve the need for perceived functional safety ASILD. Wu Junjie introduced.
Similarly, intelligent driving path planning has extremely high requirements for safety, which involves a lot of mathematical calculations, which requires the high computing power of the CPU to do support, which is embodied in the SOC, that is, ARM's A core.
At present, the A cores of the mass-produced chips on the market are independent, and do not have functional safety functions, let alone provide the lock step function contained in the functional safety core.
At present, its intelligent driving domain products have been deployed on first-class domestic brand mass production vehicles. At the same time, the newly launched cross-domain product integrates intelligent driving and intelligent gateway, which can support the all-in-one solution of cabin and berth, and has also completed the prototype stage.
Compared with the previous generation, the number of sensors that can be connected to the DCU3.0 has increased significantly, and it can now be connected to 10 cameras, 6 millimeter wave radar, and 3 lidar.
In addition, DCU 3.0 has different requirements for software platforms than traditional domain controllers.
The latter usually has only one AI chip, one MCU, and only a simple communication protocol architecture, but cannot carry the software architecture of SOA. The new generation of domain controllers is connected by several chips in series, the throughput is huge, and the transmission between chips must be carried out in accordance with the gigabit Ethernet transmission protocol, and the complexity of the deterministic communication and deterministic scheduling parts is several times that of traditional domain controllers.
To this end, Inchi Technology has deployed the protocol stack of TSN time-sensitive network in each chip to ensure its low latency and intelligent bandwidth mediation. Provides the strongest guarantee for SOA services.
"Compared with traditional domain control, cross-domain DCUS3.0 will be more uncertain about the application, so we have deployed a complete set of tool chains behind the matching software platform EMOS to support the configuration of deterministic scheduling and task scheduling, as well as the implementation of pre-processing architectures and software packages, which greatly simplifies the development difficulty of upper-level applications." Wu Junjie introduced.
Wu Junjie said that the company will continue to iterate domain controller products with stronger integration capabilities, and at the same time open up ecological cooperation with more friendly businesses to accelerate higher levels of automatic driving.