A brief analysis of Weilai’s next-generation electronic and electrical architecture

In the past two years, various OEMs have released their own next-generation electronic and electrical architectures on large-scale occasions such as technology days, such as Xiaopeng’s X-EEA3.0 central computing platform + regional control architecture, the ideal LEEA4.0 regional architecture, GAC Egypt Ann’s central computing platform architecture – Protoss architecture, Great Wall’s computing platform architecture GEEP3.0, Volkswagen’s three-domain architecture, etc.

Figure 1 Central computing platform architecture of SAIC, GAC, and Great Wall (source network)

NIO’s next-generation electrical and electronic architecture

Like other OEMs, the purpose of EEA update is to understand the problems of uncommon software and hardware, non-iterative software, and solid application functions, and to achieve software reuse, rapid iteration, rapid and flexible deployment, software platformization, and high code cohesion and low coupling. .

Figure 2 Key points of architecture update

In the next-generation electronic and electrical architecture, NIO’s choice is the regional controller architecture. The entire system consists of the central computing unit, regional controller and high-speed Ethernet to build the basic skeleton. Through the cooperation of the three, an adaptive and self-learning system is constructed. And provide a wide range of intelligent interconnection functions.

As shown in the figure below, the ring topology supports the fail-operational dual-redundancy system and has good scalability.

Figure 3 NIO’s next-generation electronic and electrical architecture

Among them, the central computing unit has a computing power of more than 1000TOPS in terms of performance, and the main frequency is greater than 1GHz. Combined with the previous information on NVIDIA’s official website, the main control of the central computing unit is NVIDIA’s Orin chip, which is iterated from the existing Adam supercomputing platform.

Figure 4 NVIDIA official platform information

Because the latest ES7 will be equipped with Weilai’s Aquila Weilai super-sensing system. There is an Adam supercomputing platform in the system. The platform consists of four DRIVE Orin system-on-chips (SoCs), providing more than 1,000 TOPS of computing power. The first two SoCs are responsible for processing 8 gigabytes of data per second generated by the vehicle’s sensor array. The third Orin acts as a backup, ensuring that the system can operate safely in any situation. The fourth Orin supports local training, improving the vehicle through fleet learning and personalizing the driving experience according to individual user preferences.

Figure 5 Weilai’s adam supercomputing platform

The Adam supercomputing platform can drive various intelligent functions and leave enough headroom for adding new functions via OTA in the future.

For the regional controller, it is an automotive cerebellar system that supports the central computing platform, and its functions are:

Distributed edge computing framework;

Vehicle Control Arbitration Center;

Information communication network for service communication;

Regional centralized data center;

Vehicle distribution hub;

Obtain data from smart sensors and control the behavior of smart actuators;

In order to realize the above-mentioned functions and realize the above-mentioned characteristics of platformization, high cohesion, low coupling, high reusability, and flexible deployment, the software must be a service-oriented architecture and widely adopt middleware (SOME/IP, DDS, etc. ).

This requires the reconstruction of the existing software architecture. The first is to introduce the remote calling method (RPC), which is also the core of the SOA idea. The second is to reconstruct the traditional and commonly used local calling method (LPC). The method is the local call of the function, there is no unified mechanism and standard), such as unifying the interface between RPC and LPC, and letting LPC simulate the communication methods such as Method, Request/Responce, Fire/Forget, Event, etc., so as to realize the insensitive call of the service to the lower layer.

In addition, due to the limited resources of the regional controller, in order to achieve cross-domain function integration and cross-domain function isolation. NIO’s approach is to use the AMP multi-core architecture on the regional controller. The RTOS in AMP mode runs an operating system instance on each CPU, and these operational instances are not necessarily identical.

Figure 6 AMP mode

This method has the characteristics of not requiring MMU, low operating overhead, and supporting system integration of different functional safety levels, but it also brings some challenges, such as the lack of unified management of core loads, the overhead of multi-core communication mechanisms, and the startup of different cores. Timing is difficult to manage, etc. With the accumulation of experience and time, this kind of engineering technology problem is definitely no problem. Besides, NIO has accumulated experience in self-developed domain control.

Figure 7 Self-developed chassis domain controller on NIO ES8

As mentioned before, whether it is a new force or a traditional OEM, they have carried out drastic innovations to their own EEA architecture in the past two years, all aiming at the central computing unit architecture, and basically they can be implemented on mass-produced models next year and next year. , in terms of the advanced nature of the architectural concept and the advanced nature of the main control chip of the main controller, it is remarkable.

However, the ability to control the advanced architecture, the ability to release performance, and the software skills are the real tests for each OEM, and this is just the beginning. The control and application of the architecture requires constant exploration, running-in, and iteration. This is the top priority of each OEM in the next stage.