On March 13, 2024, at the 5th Software-Defined Vehicle Forum 2024 and AUTOSAR China Day, Ke Zhuliang, founder & CEO of Shanghai Lingnian Technology Co., Ltd., talked about the following features of the next-generation automotive operating system:
First of all, security, which is an indispensable cornerstone. Second, scalability is crucial. OEMs expect a platform that is not limited to a single vehicle model, but can be flexibly adapted to multiple vehicle models, and even cross-border applications, to achieve rapid expansion.
Secondly, service-oriented is the direction of future development. We will gradually move from the existing signal-oriented mechanism to a full-fledged, service-oriented architecture. While this may take years, we will be able to move towards a fully software-defined vehicle pre-control architecture, or SOA, when the HPC platform architecture is truly implemented.
Finally, openness is also a core element. Openness here does not mean that all software is open sourced, but that OEMs, suppliers, and even users should participate in building a complete operating system while respecting their respective business models. This spirit of openness and collaboration will help us move together towards a more advanced era of automotive software.
Ke Zhuliang | Founder & CEO of Shanghai Lingnian Technology Co., Ltd
The following is a summary of the speech:
The evolution of the EE architecture and OS
The evolution of the EE architecture and operating system is well known. It's important to see how software evolves and iterates as architecture changes. Before 2000, automotive electronics mainly relied on embedded OS, which only processed simple signals, and was mostly used in MCUs and embedded RM cores. However, at that time, the portability of the application layer software on different chips was extremely poor, which became a major pain point.
Subsequently, the CP architecture, i.e., the concept of Classic AUTOSAR, came into being, unifying API standards and introducing RTE middleware to realize the decoupling of software, hardware and vertical levels.
Today, automotive systems are evolving to the third generation. Due to the large amount of computing power required for complex functions such as autonomous driving and intelligent cockpits, A-core and GPU chips were introduced. We are in the era of semi-SOA architectures, and although the concept of adaptive AUTOSAR has been introduced, most systems still rely on traditional APIs, forming a hybrid architecture.
Source: Speaker material
At present, we are facing two major dilemmas: insufficient communication bandwidth, and competition between OS and middleware provided by chip manufacturers. Mainstream OEMs such as Bosch USP, Xpeng SEPA, Geely EEA 2.0, Volvo SSP, etc., are all in this era.
In the future, we look forward to the fourth-generation OS based on the HCP architecture, which will realize the application layer based on atomic services. This will follow the 3A principle, where anyone, any operation, can see the required services, marking the implementation of a truly complete SOA architecture. At present, Geely, NIO and other cutting-edge OEMs are committed to researching the next generation of electronic and electrical architecture and software systems, moving towards a new stage.
Looking back on the decade of software development, the demand for intelligence has become the core driving force, leading us to realize innovative functions such as intelligent cockpit and autonomous driving. From simple parking warnings to today's AVP valet parking, intelligence is increasingly meeting the growing needs of users.
The "second half" of the EE architecture of the whole vehicle is safety and cost reduction
Over the next 10 to 20 years, software architecture will focus on two key dimensions. The first is safety, and we must create safer systems for users, especially in autonomous driving and assisted driving, to ensure that safety features are realized, so that users can truly trust and enjoy intelligent functions.
Secondly, cost optimization is also crucial. In a highly competitive market, OEMs are eager to achieve cost savings in architecture. How to help OEMs save costs in the vehicle architecture and bring economic benefits to users at the same time is the focus of the next-generation software architecture.
Although the first generation of intelligent domain control technology has been applied as proposed by Bosch, the actual demand for intelligence has not met expectations. By 2025, it is expected that no more than 10% of L3+ autonomous vehicles will actually be on the road, and the market share of L2 functions will only be about 20%.
Therefore, we should focus more on the two major needs of safety and cost. Taking safety as an example, the first consideration for users when buying a car is active and passive safety. For example, if a brand advertises a vehicle that can safely drive over iron blocks at high speeds, this safety is exactly what users expect. To achieve this kind of safety, we need to start with a methodology, including functional safety analysis, to ensure that minimal damage is achieved in a variety of scenarios.
In terms of data security, that is, cyber security, in recent years, with the improvement of the intelligence of intelligent networked vehicles, their potential risks have also attracted much attention. U.S. Commerce Secretary Raimondo's remarks about China's ability to control millions of vehicles in an instant underscore the importance of cybersecurity. How to ensure that these powerful functions are not misused is the core issue of cybersecurity.
As for fail-safe, there were early projects that aggressively pursued one-step implementation, but the high cost made it difficult for this solution to be accepted by the market. Today, we focus more on providing users with minimal security in the event of software and hardware exceptions. For example, in autonomous driving, when a sensor or chip fails, the vehicle is safely parked on the side of the road or smoothly at high speeds. This requires us to apply methods such as software backup and functional safety analysis to ensure functional safety in the event of failure.
Cost considerations are equally important. In recent years, Tesla's prices in the Chinese market have fallen sharply, and there is a significant gap in the profit margins of domestic OEMs. In the face of domestic and foreign competition and industry involution, OEMs must think about how to reduce costs in architecture.
Source: Speaker material
From an architectural perspective, there are many ways to reduce costs. For example, the Ethernetization of the entire vehicle is a hot topic. Taking Daimler S as an example, its line speed cost accounts for a large proportion of the vehicle cost. New architectures such as Volkswagen and BMW are actively exploring Ethernet to connect the entire vehicle architecture via two wires, resulting in significant cost savings for CAN, LIN buses and connectors.
In addition, node computing power rebalancing is also key. At present, there are a large number of MCUs in the whole vehicle, and their unit computing power cost is much higher than that of SOC. Considering the risk of the MCU supply chain, the MCU computing power requirements should be reduced and the computing power of the SOC should be fully utilized to save chip costs.
When reviewing the features of the next-generation OS, it is not difficult to see that in addition to emphasizing security as a core element, there are several key aspects worth paying attention to.
Source: Speaker material
The first is scalability. OEMs are looking forward to a platform that can serve across models and countries, and can quickly adapt to different needs and achieve flexible expansion.
Secondly, service-oriented architecture is an important direction of our transformation. We are moving from a traditional signal-oriented mechanism to a complete SOA architecture. In the future, when the HCP architecture is realized, we will have a complete domain control architecture for software-defined vehicles to achieve a more intelligent and service-oriented automotive ecosystem.
Openness is also indispensable. What we mean by openness is not simply open source software, but refers to the fact that in the process of software development, OEMs, suppliers and even users should participate together to build a complete OS system and achieve wider cooperation and innovation.
In terms of security, we have made in-depth arrangements in key areas such as communication and scheduling. How to ensure the accurate scheduling of complex tasks between the MCU and the SOC is the top priority in the SOA architecture. At the same time, we are also strengthening our global security-related OS services to provide solid support for embedded systems, chip drivers, and time synchronization.
Deterministic scheduling
The concept of scheduling and communication may not be new, but its importance in life cannot be overlooked. In the case of the transportation system, for example, the precision and reliability of the subway is in stark contrast to the disorderly congestion of the highway system. Similarly, in automotive software systems, we strive for precise execution at every step to ensure that emergency functions such as AEB can function reliably and reliably. By accurately breaking down each step, from signal input to task processing, we ensure the smooth and efficient calculation chain, providing a strong guarantee for the intelligent and safe driving of the car.
Source: Speaker material
There are three main types of scheduling: time-based scheduling to ensure that tasks are executed on time, data-based scheduling to process data such as LiDAR data to arrive, and event-based scheduling to ensure the continuity and consistency of the computing chain such as perception, planning, and control.
In the current communication architecture, MED's architecture is interwoven with multiple networks, including complex bus systems such as CAN and LIN. Is it possible that these buses will be integrated into the Ethernet of the whole vehicle in the future? Many people still have reservations about the reliability of CAN networks as a means of data transmission. So, how do you secure communications while pursuing unification?
This brings us to the concept of the transmission of time. It is not new to friends who have learned communication. Essentially, it is similar to TDMA, the concept of instant multiplexing. We applied it to in-vehicle communication, and realized the hybrid orchestration and efficient transmission of various signals through accurate time-slice multiplexing. In the early applications of TDMA, the vocal signal was transmitted as a time slice at 4.1 kilohertz. Now, we are applying it to automotive Ethernet to ensure reliable transmission of vehicle data.
This communication product has been tested in the laboratory, and the experimental data are concentrated in a very stable and narrow range, which fully proves its certainty and reliability. In fact, this technology is not only used in the automotive sector, but also in aerospace and other non-civilian industries.
So, what can these products bring us? First of all, they are the key to achieving high-level intelligent driving and ensuring driving safety. Secondly, through the application of technology, we are able to achieve lower latency in the central gateway system. In addition, by putting these applications in the SOC, we can increase the load capacity and reduce the dependence on the MCU computing power, thereby saving costs. Finally, the popularization of Ethernet in the whole vehicle will greatly reduce the use of wiring harnesses and improve the manufacturing efficiency of the whole vehicle.
Here, I would like to give a brief introduction to our company. Founded in 2021, Lingnian Technology focuses on intelligent driving platform software, the underlying and middleware software technology with completely independent intellectual property rights, builds the interconnection between the underlying data and the application platform with high reliability, high security and high real-time, and provides a completely autonomous and controllable operating system, a highly scalable security basic software platform and customized solutions for intelligent vehicles and intelligent driving.
(The above content is from the keynote speech "Building a Highly Secure and Reliable Operating System" delivered by Ke Zhuliang, founder & CEO of Shanghai Lingnian Technology Co., Ltd., at the 5th Software Defined Vehicle Forum 2024 and AUTOSAR China Day on March 13, 2024.) ）