Why is the self-driving industry paying attention to operating systems two or three years later than chips?

Friends who continue to pay attention to the self-driving industry may find an interesting phenomenon: the topic of operating systems is far less popular than chips and lidar. Even, the industry's attention to the operating system is at least two or three years later than that of chips - chips are already a hot topic of "out of the circle" in 2017, and the operating system, until 2019, is still relatively "unpopular" in the circle.

Why is this difference?

With these questions in mind, in November, Nine Chapters of Smart Driving interviewed Bi Xiaopeng, co-founder and vice president of R&D of HuayutongSoft, Shang Jin, general manager and CTO of Sinoc Motor, Wang Haowei, chief architect of Zhongke Chuangda, Mao Haiyan, director of Neusoft Ruichi's business line and general manager of global sales in Europe and the United States, and many other experts in the field of operating systems.

The consensus of the respondents on the answer to this question is as follows:

This is similar to the development process of smart phones, we first hope that their products have obvious differences with competitors in terms of function, and hardware is the basis and guarantee for achieving new functions (such as traditional chips can support vehicle control, but can not be used for automatic driving), therefore, in the early stage of automatic driving, chips (mainly the size of computing power) and other hardware has won high attention. In contrast, the high-performance operating system not only solves the problem of "can not", but also "good or bad", and the early automatic driving function is relatively simple, generally do not need a high-performance operating system.

L1-L2 automatic driving, often in the distributed EE architecture to achieve, each functional algorithm corresponds to an independent chip, and these functional algorithms are relatively simple, naturally do not need high-performance chips, MCU (single-chip microcomputer) can be. These MCUs run RTOS such as the Autosar Classic Platform (Autosar CP). Autosar CP is equivalent to a complete operating system including kernel + middleware, and although the functional level is low and the limitations are many, it seems to be "enough".

Simply put, the traditional ECU undertakes limited functions and relatively simple operation, such as sensor processing only involves simple, low-resolution point clouds or digital images, etc., often only need to schedule a single task, so there is no need for a high-performance OS to achieve resource scheduling and allocation.

More critically, at this stage, the functional algorithm and the MCU are highly coupled, the MCU bought by the main engine factory from the supplier, often comes with its own algorithm, the main engine factory only needs to do a good job of integration work, no need to write its own algorithm, therefore, naturally there is no incentive to pay attention to the operating system. The biggest "customers" are not in a hurry, so where does the motivation of entrepreneurs and investors come from?

In contrast, from L2+ upwards, functional algorithms have become very complex, distributed EE architecture can not meet the needs, must be based on centralized EE architecture to do development, and in the centralized EE architecture (domain controllers, central computing platform), a variety of algorithms run on a system-level chip (SOC) similar to Xavier, a large amount of information needs to be collected to domain controllers, and then sensing, computing, decision-making, data transmission and fusion, much more complex than in the distributed architecture. In this case, the amount of data that the SOC chip needs to process is growing exponentially, and the amount of tasks involved is also more, and many tasks are scheduled simultaneously by multiple threads, which requires a strong enough real-time operating system to better allocate, schedule operations and storage resources.

In addition, although the SOC and MCU play a similar role, the former is much more difficult to achieve functional safety (algorithm security, real-time data, etc.) - MCU is only suitable for simple scenarios with clear requirements and low requirements for computing power and latency, correspondingly, ROTS running on it only needs to provide some scheduling mechanisms and process some simple logic; and intelligent driving SOCs need to deal with complex scenarios, the system's algorithms are complex, rich in functions, and dynamic deployment. Autosar CP has been unable to meet the requirements, and a high-performance operating system has gradually become a "just need" for developers.

The more important reason is that whether it is Tesla, Wei Xiaoli, or other domestic OEMs, or Tier 1, everyone's hardware framework has basically been determined, and the functions that can be achieved based on hardware have been basically determined, no matter who the chip is used, the functions that can be achieved are basically image perception, etc. Therefore, everyone naturally focuses on the differentiation of the software when pursuing differentiation.

When everyone begins to pay attention to software differentiation, the decoupling of software and hardware will become an inevitable choice, regardless of whether there is ability or not, all enterprising OEMs are no longer willing to accept the "software and hardware integration solution" provided by suppliers. In the context of "software-defined cars", OEMs must sink down, develop their own software algorithms, be able to define functions, develop differentiated applications, and achieve OTA upgrades, otherwise they will miss the most important value in the entire life cycle of the vehicle. And to do this, you can't help but focus on the operating system.

Many respondents pointed out that from the history of the smart phone industry, in the early stage of industrial development, it is usually the operating system that follows the chip manufacturers, and when the whole machine factory (mobile phone factory manufacturer, automobile manufacturer) buys the chip, the operating system has been "matched" - at this stage, how to choose the operating system is more of a problem that chip manufacturers need to pay attention to, and the downstream machine factory does not need to "worry too much"; but with the improvement of product maturity, the importance of the operating system is also highlighted step by step.

From the perspective of the demand for automatic driving, the chip solves the problem of "can not", while the operating system solves the problem of "good or bad". Until the question of "whether it is possible" has not been solved, "whether it is good or not" is naturally not an urgent issue; and once the difficulty of "whether it can be" is broken, "whether it is good or not" will naturally rise to "the main contradiction at present.".

In short, in the early days of the development of the autonomous driving industry, chip manufacturers flew with operating systems; but as the industry gradually entered a mature period, the operating system's voice will become more and more powerful. Many respondents believe that at a later stage, when the industry development is highly mature and developers are highly concentrated, the market concentration of the operating system will be very high.

The narrow operating system mainly refers to the kernel of the operating system, while the generalized operating system also includes the "middleware" above the kernel and under the application software. In april this year's "status quo and future of the automatic driving OS market" article, the author focused on the analysis of the core of the automatic driving OS, in the next one or two months, the author will use 3-4 long articles on automatic driving middleware to do a "popular science", interested friends please keep an eye on our update.

Write at the end

Communicate with the author

If you want to communicate directly with the author of the article, you can directly scan the QR code on the right and add the author's own WeChat.

Note: When adding WeChat, be sure to note your real name, company, and current position

As well as information such as intention positions, thank you!

Industry Transformation Exchange Group

To exchange difficulties encountered in the transformation and share the experience in the transformation process, you can scan the QR code on the right to add the staff WeChat, provide a business card and personal resume, and then pull you into the group.

About submissions

If you are interested in contributing to the "Nine Chapters of Intelligent Driving" ("Knowledge Accumulation and Collation" type article), please scan the QR code on the right and add the staff WeChat.

"Knowledge Accumulation" manuscript quality requirements:

A: The information density is higher than the vast majority of reports of the vast majority of securities companies, and it is not lower than the average level of "Nine Chapters of Intelligent Driving";

B: Information is highly scarce, more than 80% of the information needs to be invisible in other media, and if it is based on public information, it needs to have a particularly strong and exclusive point of view. Thank you for your understanding and support.