SOA architecture helps the implementation of the travel and berthing integrated domain controller

As a Tier1 company with strong software genes, Neusoft Reach believes that on the basis of good hardware architecture design, doing a good job in hardware adaptation layer is the key to decoupling software and hardware, and it is also the core of improving R&D efficiency and shortening product delivery cycle. The following is a summary of the speeches:

Dr. Jun Hu | Deputy Technical Director of Neusoft Reach Autonomous Driving Business Line

The vehicle architecture has gradually evolved from the traditional distributed ECU to the centralized domain, which is the change that our automotive industry, especially the autonomous driving industry, has been experiencing in recent years. Many partners, including Neusoft Reach itself, are also thinking about how to do a central computing architecture and how to carry out cross-domain centralized work in the future.

On the surface, it is the change of the vehicle architecture, combined with the trend of the development of autonomous driving functions and the trend of "software-defined vehicles" gradually landing in the industry, in fact, this process is very powerful for the entire industry to reshape, our remodeling boils down to three levels of reconstruction:

The first is the restructuring of the entire industry's organizational structure. Whether from the perspective of Neusoft Reach Tier1 or the perspective of OEMs, in order to develop more complex driving and parking integration functions, our organizational structure has been adjusted accordingly - the proportion of software teams is increasing, and more and more barriers are being pulled down by the original vertical division by function... We've all gone through this process from all angles. From a technical point of view, it is believed that there may be many deep convergence between domains in the future. But why haven't we seen these deeply integrated product forms emerge in a short period of time? More because the inherent organizational structure cannot be well adapted to the construction of new products, so the restructuring of the organizational structure will continue to occur in our industry.

The second level of reconstruction is the reconstruction of product form. Since Neusoft Reach entered the industry, it has found that the products needed by the mainstream market in the past are simply function-oriented and take SOP as the end node of R&D. Looking back today, a lot has changed. From the perspective of software technology architecture, the market requires us to have more openness, so we will work with customers and third-party partners to create relevant products during the research and development process. At the same time, product complexity is also increasing, and we need to continuously give products the ability to iteratively update throughout the life cycle of the product, which requires us to prepare for product architecture and software architecture.

The third level of reconstruction comes from the reconstruction of the industrial chain form. The traditional industrial chain is more of a vertical "chimney" closed industrial structure, and now this structure has been broken. The current industrial chain cooperation mode is no longer the traditional chain structure, but a more flat network structure. In the process of landing complex domain controllers, there will be many participants, and the research and development process is the process of you and me, and I have you.

At the same time, market factors play a role in reshaping the future product form and technology form. First of all, the installation rate of autonomous driving and intelligent driving functions has undoubtedly reached the explosive period, and in the next 2-3 years, the mainstream functions of automatic driving should be integrated in the L0-L3 range. By 2025, 90% of its functions will be concentrated in the L0-L3 range, but does this mean that functions will be standardized in the next few years? In fact, this is not the trend implied by the market.

More and more of our customers and partners pay more attention to the ability of personalized development and iterative improvement of functions, how to expand the scenarios required by customers in the process of iterative improvement of functions, and how to effectively use technology to support this matter, which is also a problem that we need to think about in the research of autonomous driving domain controllers.

Another trend is cross-domain integration, my view is basically the same as that of Gascar Automobile, cross-domain integration is not a state of full integration in the short term. We have seen a lot of trends and needs for cross-domain integration from the functional level, and the future deep cross-domain integration will definitely occur, so what should the future technical architecture look like? This is the demand and task that the market puts forward for us.

Based on these understanding and judgment of the market and industry, we believe that to develop autonomous driving domain controllers, we need to face these challenges at the technical level: first of all, the sensor level, when faced with the jump in the number and type of sensors, how to maximize the use of sensors in high-end complex autonomous driving functions, and do a good job in the strategy of sensor fusion according to scenarios and functions, which we need to consider.

In addition, as a Tier1 supplier, Neusoft Reach will encounter sensors with different technical parameters provided by different suppliers in different projects, in this case, how to generalize sensors and realize the reuse of core software assets is also a problem we need to consider.

With the jump in the functions of domain controllers, as well as the increase in hardware complexity and software complexity, our R&D investment and R&D difficulty are also increasing, and the market requires us to achieve delivery in a shorter cycle, which requires us to do a good job of software separation. From the perspective of software module and architecture design, we need to do a good job of abstraction, layering, centralization, and integration as much as possible, and integrate from distributed to centralized. In addition, in the face of open market demand, how to do a good job of cooperation, white box, gray box, black box and even hybrid delivery work also need to be considered.

Moreover, the complexity and iteration speed of the current hardware have been greatly improved, and how to do a good job in adapting and designing at the software level to maximize the hardware efficiency is also a problem that needs to be considered.

Finally, for a complex autonomous driving domain controller, the landing of functional safety and information security at the product level has undoubtedly become the consensus of the industry.

In the past, when organizing the development of an autonomous domain controller, we were more based on the structure of functionality and vertical segmentation of product lines. In the era of domain controllers, we need to break the boundaries and barriers between traditional functions as much as possible, and use a more three-dimensional and more sophisticated R&D team to develop complex domain control products at the application level.

After the change of R&D model, we need to think about what is our position in the era of software-defined vehicles, as an autonomous driving domain controller supplier, and what are the core points we should pay attention to in the process of change.

We believe that more core capabilities are reflected in good hardware architecture design, and doing a good job in hardware adaptation layer is the key to our decoupling of software and hardware, and it is also the core of improving R&D efficiency and shortening product delivery cycle. As a supplier with strong software genes, Neusoft Reach has a core capability in building a software ecosystem based on basic software and middleware, which includes these levels:

First of all, NeuSAR, Neusoft Reach Automotive Basic Software Product, will be iteratively updated to version 4.0 based on AUTOSAR, and on this basis, we will provide generalized and SOA-based intermediate components to all vehicle domains.

For different autonomous driving domains and BMS products, we also provide functional middleware for a specific purpose. Based on these interface foundations and services, we will build our own with a more open mind, or work with partners to build personalized and flexible upper-layer applications.

What characteristics should a software architecture that adapts to today's autonomous domain controllers and future cross-domain central domain controllers have?

First of all, there is no doubt that SOA must do its job well. It includes communication-oriented service component functions and service-oriented software architecture service work, on the basis of which we do a good job in the separation of software and hardware, as well as the decoupling and separation of the underlying software and the upper-layer application software.

Self-evolution is also a feature that products need to support. How to support more scenarios in the follow-up of the product, how to maintain the iteration cycle of the product, and how to develop the commercial value of the product is undoubtedly a very important mechanism. There are three points of this mechanism: first, the vehicle side needs to support the mechanism of value data discovery at the level of software architecture; In the cloud, it is necessary to have the support of vehicle-cloud interaction, and realize the support of products and software and hardware in the cloud; At the same time, we need to have a good OTA mechanism to help us continuously update the software.

Functional safety and information security are also capabilities required for product implementation. Finally, how to do a good job in cross-domain SOA service frameworks and components for the future is also a key issue that we need to consider in the existing software architecture and technical architecture.

Based on the above considerations, Neusoft Reach refers to the domain control architecture and logical architecture diagram, first of all, we will abstract the performance domain and security domain of the underlying hardware computing resources, and equip the basic software and middleware components of Neusoft Reach for the performance domain and security domain, corresponding to the implementation of AUTOSAR respectively.

On this basis, in response to the requirements of the self-driving domain, we further abstract, encapsulate and isolate the underlying hardware resources through the runtime environment abstraction layer in the software architecture, and expect to achieve software and hardware separation and software abundance.

In the process of landing time of related products, we found that in order to better realize decoupling and realize the concept of SOA, we need to have a good communication framework and communication middleware. Core components are unified packaging for the operating system and self-driving core operating environment. The module module encapsulates Neusoft Reach self-developed key capabilities of autonomous driving, including core capabilities such as perception, positioning, and decision-making. The Framework provides a good SOA mobilization framework and a full lifecycle management component of system software entities. On top of these components and architectures, we build servitized applications on top of it.

When developing the application layer of a domain control product, there will always be a choice between the communication protocol and the corresponding underlying communication link. However, on different domain-controlled hardware platforms and in the face of different customers, we will find it difficult to achieve the same application and the same data flow. How to ensure good reusability and migration of the application layer? This requires us to realize service-oriented encapsulation of communication middleware and provide standardized interfaces to the application layer. Such a set of packaging and communication middleware has been integrated in the latest version of Neusoft Reach products, which can carry out intra-domain or cross-domain work based on Some/IP and DDS, and realize relatively standardized and configurable communication service interfaces.

In terms of functional safety, Neusoft Reach has obtained the corresponding functional safety process certification, and in the actual product development process, our functional safety has also been actually implemented in the domain control platform project. In terms of information security, Neusoft Reach has corresponding experience and software product capabilities in firewall, secure boot, secure storage and other levels, and has become the first batch of enterprises to pass ISO 21434 process certification.

Based on such a set of software architecture, Neusoft Reach has launched a series of products, including DMS products that support functional safety, 1R1V products equipped with SOA software architecture, and based on a relatively open architecture, in this product and partners through joint research and development, to achieve the final landing of the product.

Our L2+ integrated domain controller is also equipped with SOA software architecture and will achieve SOP by the end of this year, with sensors currently supporting 8 million cameras and a total of 11V sensors. Typical supported features include typical L2+ features such as TJP, HWP, and HPP.

In addition, the corresponding project of our cross-domain convergence central computing unit has also been launched, and we are currently advancing research and development with heavyweight partners. We will launch this product in the market next year.

Since Neusoft Reach entered the autonomous driving industry, it has accumulated a lot of experience in the development and implementation of autonomous driving hardware platforms. At the software stack level, we have our own basic software and suites, as well as middleware software assets and services for autonomous driving. We have standardized packaging and decoupling for various types of sensors, and packaged the core competencies required for autonomous driving. In the process of landing each project, we have the ability and software assets to implement L0 to L3 autonomous driving functions, and this year we have also landed Neusoft Reach self-developed intelligent driving cloud, which can cooperate with the vehicle-end platform to achieve closed-loop product research and development.

Based on these development experience, software assets and industry capabilities, Neusoft Reach is willing to uphold a more open and cooperative mentality, and our partners continue to launch more products with user value for the Chinese market and users, as well as future-oriented central domain control convergence products.

(The above content is from the keynote speech "SOA Architecture Helps the Landing of Integrated Parking Domain Controllers" delivered by Dr. Hu Jun, Deputy Technical Director of Neusoft Reach Autonomous Driving Business Line, hosted by Gesch Automobile, guided by the Administrative Committee of Shanghai Hongqiao International Central Business District and the People's Government of Shanghai Minhang District, and co-organized by Shanghai South Hongqiao Investment and Development (Group) Co., Ltd. on November 14, 2022.) ）