Analysis of future automotive electronic structure under the demand of AI and autonomous driving

This is the electronic and electrical architecture of a mid-size luxury car, as shown in Figure 1. It is a parallel network composed of different bandwidths and signal transmission standards (protocols) connected by gateways. There are CAN (controller area network), FlexRay, MOST, LIN, LVDS (Low voltage differentiable signaling), and even direct connection of analog signals.

Figure 1 EE architecture of a luxury car

The in-vehicle electronic network already has tens of thousands of signals sent and received under different transmission standards (protocols), and different forms of software are running in hundreds of different electronic control modules. The scale of software has exceeded 100 Mio. lines. How to control such a complex system, effectively verify its overall function and safety, and at the same time control the R&D and vehicle costs within a certain range, this is a huge challenge that OEMs are facing now.

Let's review the development history of personal computers. Since IBM invented the PC in August 1981, the hardware architecture of computers has been basically fixed and scalable, with CPU, GPU, PCI-Bus, ports USB, VGA, HDMI, working memory, hard disk or SSD; in terms of software, there is an operating system, different application software, and the network connecting these PCs is Ethernet.

Figure 2 IBM PC5150

Here the author points out that there are huge differences between automotive electronic modules and PCs:

Figure 3 Comparison between automotive electronic modules and PCs

In the face of the vehicle-to-everything (V2X), ADAS (advanced driver assistance system), vehicle electrification, and the future autonomous driving market, the existing automotive electronic and electrical architecture can no longer meet future needs. A large amount of data (signals from radar, lidar, cameras, and other sensors, as well as V2X signals) needs to be received, processed, and corresponding instructions issued to the actuator.

In this regard, the author puts forward the following three viewpoints:

First: Develop and formulate a standard, scalable computer for automobiles, called iPC (in-vehicle computer)

iPC is like a high-performance, scalable PC, with standard CPU (central processing unit), GPU (graphic processing unit), SoC (system on chip), and other scalable AI chips, such as NPU (neuronal network processing unit). In 2000, when I was in charge of powertrain electronic systems at the German Daimler Group, I had the idea of ​​iPC when I was discussing future automotive electronics with my colleagues.

The operating system of iPC can be unified, such as Linux. The network port of iPC is mainly Ethernet, with a bandwidth of 1Gbps to 100Gbps, integrating the future 5G mobile network and I/O ports. This is a once-in-a-century opportunity for PC and server manufacturing companies and innovative technology companies! It is also a time for the automotive electronics industry to reshuffle!

Second: Develop and implement automotive Ethernet

Ethernet has been widely used in the Internet, personal computers, and industrial automation. The future China Manufacturing 2025, or Germany's Industry 4.0, and the Internet of Things (IoT) all have a common carrier, Ethernet! One of the main obstacles to the fact that Ethernet has rarely been used in automotive networks (except for a few multimedia applications) is the real-time capability of Ethernet.

Ethernet generally has to wait for the first transmitted signal package to be delivered before the second signal package can be transmitted. In this way, the control signal of the engine, steering or braking system cannot meet the requirements. In order to enable Ethernet to be controlled in real time (time trigged), IEEE802.3 has established standards to send express signal packages. It can pause the signal package in transmission first, let the express signal package arrive in real time, and then transmit the paused signal package to solve the real-time control problem of the engine, steering and braking system. In addition, if PoE (Power over Ethernet) is used, some actuators can be driven directly without adding external wiring harnesses.

Third: Develop application software independent of hardware

There are a wide variety of existing automotive electronic control modules. Because they come from different Tie1s, have different hardware, different operating systems, and even different application software, this is an extremely costly, high-risk, and very complex integration and verification task for OEMs.

If we have an in-vehicle computer iPC and a unified operating system for iPC, it will give software technology companies a huge opportunity to develop hardware-independent application software, just like App, which can run normally in different hardware environments. Customers can even pay to download according to their personal needs (new business model), and OEMs can easily integrate, verify, and reuse (re-use). Some of my colleagues in the German automotive mainframe factory (BBAP) also hope to have hardware-independent application software to facilitate further development, re-application (re-use), and even switch Tie1.

Figure 4 Developing hardware-independent application software

Figure 5 Summary of automotive electronics based on iPC

The future architecture of automotive electronics will be in-vehicle (automotive ethernet).

The author makes three points here:

1) Standardized and scalable vehicle-mounted computer iPC

2) Unified operating system (such as Linux)

3) Hardware-independent application software

This will be the trend of future automobile development, bringing new business opportunities to PC and server manufacturers, technology innovation companies, and independent software technology companies. Breaking the monopoly of system Tie1 is the century-old opportunity for Chinese parts companies!

It is predicted that by 2025, the penetration rate of Ethernet in automotive electronics will exceed 50%, which will not only bring about a technological leap, but also new breakthroughs in cost reduction, lightweight and car personalization (iPC App, software update OTA), vehicle networking V2X, and unmanned driving (routing & map via cloud).

The author, Dr. Liu Xiaoyi (founder of CostKey-Solutions.com), has 21 years of senior automotive technology experience in the German Daimler Group (Mercedes-Benz Automobile) (Senior Manager of Powertrain Electronic Control System Series Development and Commercial Vehicle Electronic Architecture Design, Vehicle Planning, Cost Engineering, New Process Development, Transmission Production, etc.), and has led the development of Mercedes-Benz Heavy Truck 16-speed Automatic Transmission Series, Powertrain Components and Airbus Large Structural Components New Process Development, Electric Commercial Vehicle Vehicle and Cost Planning, and Commercial Vehicle and Sedan Multiple Cost Reduction Key Projects (PKO/OPTIMA, CTX, CORE). He served as a senior technical manager at the Great Wall Motor Technology Center and established the first fully-established cost engineering system, methods and processes for Chinese automakers for the Great Wall Motor Group, and integrated it into the entire product process of the group.