“Accenture’s former CEO Pierre Nanterme said that if we continue to promote digital transformation, we can turn the challenge of product design complexity into a competitive advantage for our business and products. In today’s digital age , Every company in our industry, including the automotive industry, is facing enormous pressure for digital transformation, including digital R&D, digital manufacturing, digital management, and digital services. I believe this is a transformation issue that every manager must face. .
In the past ten years, Siemens has invested more than 10 billion US dollars in total, insisting on digital transformation, providing a full range of digital solutions to the industry, including the automotive industry, especially the automotive autonomous driving industry, to ensure that business partners have the last laugh in the digital wave. Win to the end. —— Huang Hanzhi
In the past ten years, Siemens has invested more than 10 billion US dollars. Through continuous product research and development and the acquisition of leading companies in various fields around the world, Siemens has formed a comprehensive range from autonomous driving chip and prototype controller design, autonomous driving system development, vehicle performance design, Electronic Electrical architecture design and software engineering, testing and verification solutions, equipment and design technology at the level of urban smart transportation, provide the most complete and leading digital solutions for autonomous driving development, testing and verification.
Mr. Huang Hanzhi is the Greater China Director of Siemens Digital Industry Software Autonomous Driving Product Line. He has extensive engineering and market experience in the fields of digital simulation development and testing of ADAS and autonomous driving systems, and simulation development and testing of crash safety performance.
Huang Hanzhi has nearly 20 years of experience in the industry. He has been engaged in the product development and application of automotive driver assistance systems and active safety systems in Delphi and Continental; then led TASS International China Branch to engage in the simulation development of driver assistance, autonomous driving systems, and vehicle crash safety performance. and testing business; with the integration of TASS International into Siemens Digital Industries Software, he continues to provide leading technology and solutions to industry partners. He also holds a number of patents related to intelligent driving.
Zosi Automotive Research conducted an exclusive interview with Mr. Huang Hanzhi.
Huang Hanzhi, Director of Siemens Digital Industry Software Autonomous Driving Product Line Greater China
Q1. After PreScan was acquired by Siemens, which Siemens simulation tools did it mainly achieve? What are the major enhancements to PreScan?
Huang Hanzhi: Let me briefly introduce the history of PreScan. PreScan is a tool for the simulation of autonomous driving systems, first developed and commercialized by the Dutch company TASS International. In China, the commercialization of PreScan started in 2011, and it has been 10 years since then. TASS International was acquired by Siemens in 2017.
As the world’s most mature, commercialized and technologically advanced digital twin tool for autonomous driving, PreScan and Siemens simulation tools are connected in the following aspects:
First, Siemens also acquired Mentor in 2017. Mentor is not only a global giant in the field of EDA, but also a leader in digital tools for software life cycle management (ALM). tool.
The connection between PreScan and ALM software life cycle management tool directly exports the system functional requirements and software requirements, and exports the test use cases through the requirements management tool, and executes them in the PreScan simulation test environment. The results of the optimization and improvement of the PreScan test can also be directly corresponding to Software version management.
Second, the tool for vehicle dynamics, Siemens has AMESim system simulation tool, which is also the most sophisticated chassis dynamics tool in the world, which is connected with PreScan.
Third, autonomous driving requires a large number of test conditions. Siemens also has an AI-based scene creation, search, and optimization tool HEEDS, which is also connected with PreScan.
Fourth, Siemens has an autonomous driving road data collection solution—SCAPTOR, which can collect road data, extract key scenarios, model them, and then import them into PreScan to enrich the test scenario library.
Fifth, Siemens’ enterprise-level platform and data management tool, Teamcenter, is one of the best enterprise-level data management tools in the industry. Regardless of the massive test cases just mentioned, demand management, or software version optimization, it can be integrated with the Siemens Teamcenter data management platform. achieve open.
Sixth, we also provide consulting and supporting services. Siemens has a closed test site for autonomous driving, a test site that complies with standard certification. In terms of serving car companies, we assist in the establishment of the digital R&D process for autonomous driving.
What features have been enhanced with PreScan? There are five main points here:
① Significantly improved scene visual effects rendering;
② Improve the real-time performance of software operation and compatibility with the world’s major real-time hardware platforms;
③ The physical model of environmental perception and sensors has always been the absolute advantage of PreScan in the field of autonomous driving;
④ Automatic creation, search and screening of massive scenes of autonomous driving;
⑤ The cloud deployment of PreScan supports cloud computing.
Q2. How does PreScan merge into Siemens tool chain, how does it help to win customer orders?
Huang Hanzhi: The biggest difference between the automatic driving system and the traditional automotive electronic system is that its product life cycle does not end with the automotive SOP. With the popularization of OTA technology, even after the autonomous driving system is delivered to end users, it will continue to optimize – iterate – upgrade – update – release. If the enterprise retains a digital twin of the autonomous driving system in-house, it can realize the optimization iteration of the whole life cycle.
The research and development of autonomous driving systems cannot be completed by a single research and development department, nor is it a separate research and development system. As mentioned just now, although Siemens PreScan is a system development tool for autonomous driving, it is related to the management of functional requirements and software version management, including the automatic creation and search of massive test scenarios, and the modelling of road data for autonomous driving. And together with the enterprise data management platform, it constitutes a comprehensive solution. In this way, customers do not need to choose digital tools from different suppliers, saving development costs and improving efficiency.
I believe that an all-in-one solution is very attractive to enterprise customers.
Q3. Siemens has realized the system development support for the closed-loop full life cycle of autonomous driving. Will it reduce the openness of PreScan? If openness is maintained, what progress has been made in openness in the past two years?
Huang Hanzhi: The openness of PreScan has made great progress in recent years.
First of all, PreScan has been a very open software platform since its commercialization, especially for partners and users of PreScan digital simulation tools, who use PreScan to test their autonomous driving algorithms, the compatibility with the algorithm platform has always been very high. open and easy to use.
Secondly, we also noticed that in recent years, technology companies and Internet companies have entered the field of autonomous driving. In order to meet the needs of these customers, PreScan has evolved from a simulation tool based on graphics and interface operations to an API interface, The function of coding operations.
Moreover, Prescan’s format compatibility for scene data including OpenX format is also very good.
PreScan is also excellent in compatibility with multiple cloud platforms. We can disclose business cooperation cases, such as compatibility with Microsoft’s Azure, Amazon AWS, compatibility with some domestic cloud service providers and cloud platforms, and compatibility with private cloud platforms of domestic engineering companies. In the real-time field of RT (Real time), it also strives for compatibility with mainstream real-time computing platforms, and has been continuously optimized and enhanced.
Q4. Who are the main customers of Siemens’ autonomous driving road data collection system? Can these collected real scenes be easily converted into a simulation scene library?
Huang Hanzhi: Siemens’ autonomous driving road data acquisition system, SCAPTOR, is a new product officially released in fiscal year 2021. Its technology sources are mainly in Germany, and the main customers initially are OEMs and OEMs engaged in ADAS and autonomous driving in Germany. parts supplier. At present, this system has been promoted in China, and great progress has also been made. The real scene collected by SCAPTOR can be easily converted into a simulation scene model, mainly including the following technical points.
First, when collecting road data, mark the collected scene data manually or automatically to mark its danger level.
Second, the collected real scene, it is planned to extract the truth value. On the one hand, it uses Siemens’ internal open technical capabilities to label and extract the truth value, and at the same time, it also cooperates with an external service company that does truth value labeling and extraction.
Third, with the labeling of these ground-truth data, critical and dangerous Corner cases will also be screened, and finally these key scenarios will be modeled and converted into a simulation scenario library compatible with the PreScan simulation tool.
Q5, PreScan and Siemens’ digital twin technology are well integrated. What are the application needs of customers for digital twins? What are the development trends and technical challenges of digital twins in automobiles?
Huang Hanzhi: At present, the domestic customer groups of our autonomous driving digital development tools can be divided into the following categories. One is a small number of leading OEMs, who follow the trend of independent research and development and software-defined vehicles, and build their own autonomous driving systems, including algorithms, software, and even computing platform design capabilities. The application of simulation tools by these companies is relatively unique, and the demand for digital twin applications covers a wide range. for example:
First, the research and development of sensor environment perception clustering algorithms in different scenarios.
Second, do perceptual fusion from the raw data level.
Third, when implementing the company’s customized ODD (design operation area), the evaluation of the configuration of environmental awareness sensors required, or the analysis of capability, combination, quantity, and availability.
Fourth, training and testing optimization of the control planning decision algorithm.
Fifth, design autonomous driving planning and control algorithms based on the physical limits of vehicle chassis dynamics, which is the most comprehensive type of customer.
The second type of customer has some limitations in the application of digital twins, which is also the situation faced by many OEM customers. They are also aware of the applications required by digital twin technology, but their own technology accumulation may still need to be advanced step by step. At present, many OEMs use digital twin tools for hardware-in-the-loop testing. Most of our hardware-in-the-loop tests focus on functional verification, while others only focus on the validation of state machine logic. The second category of customer applications will be slightly narrower.
The third type of customer is sensor companies. In recent years, many outstanding vision, millimeter-wave radar and lidar technology companies have emerged in China. These companies will use our very sophisticated PreScan environment perception sensor model for the optimization of their sensor ontology design, training, testing and optimization of target clustering algorithms, etc.
The fourth category of customers is some technology companies. They also realize that the future is to adopt an all-cloud autonomous driving digital twin solution, so whether it is a massive scene library, a system simulation digital twin platform, and the evaluation of the final test performance indicators, it must go all-cloud.
Q6. Cases of digital twin applications
Zoth: Digital twins are still relatively abstract. We look at some application demonstrations of some autonomous driving companies, there are often multiple windows, one of which is the real scene seen by the vehicle camera, and there are multiple small windows next to it, one of which is used for virtual Display Perceived pedestrians and vehicles. Is this the application of digital twin technology you are talking about?
Huang Hanzhi: I am also imagining the picture you describe. According to the scene you mentioned, the first window is the original image collected by the on-board camera, and the second window is the window for the recognition and feature extraction of surrounding traffic participants by its environment perception algorithm. I think this is a very vivid Display .
But we must know that digital twin technology is not just a visual image display. Our digital twin for autonomous driving contains a digital twin of multiple aspects, not only the traffic scene, but also the characteristics of environmental perception, including the results of algorithms, and the characteristics of chassis dynamics. All four parts must have a complete digital twin model.
This one you mentioned, I guess it may be environmental perception. The digital twin should also reflect some characteristics of the environmental perception sensor and the performance of the perception results. It is not only an intuitive representation of an image, but there are more details below the representation.
Zoth: Another application scenario is testing grounds. For example, there is a test vehicle on the test road, but the number of vehicles participating in the test is relatively small, while there are many vehicles in the real traffic scene. In the test field (due to cost and other reasons) it is not possible to run many vehicles at the same time, so many vehicles need to be virtually generated at this time. Through the combination of virtual and real, a more complete transportation system is formed. Is this also a digital twin application?
Huang Hanzhi: Your description is very good. In some of our past business practices and business projects, we are also aware of this need. But again, the presentation will give us a superficial impression that there are many application models behind it.
For example, recently we are in the field of V2X, especially in the field of intelligent driving based on V2X, whether it is to do in-vehicle communication modules, terminal stress testing, interconnection testing, protocol stack testing, or upper-layer V2X-based testing. The test of the application algorithm of intelligent driving is faced with a problem. It needs a test field with better 5G infrastructure and a target vehicle composed of other traffic participants, which is bound to pose a very big challenge to the cost of physical testing.
But we can greatly reduce the cost of physical testing and improve testing efficiency through digital twins. The method used is: the tested host car and the on-board module of the host car are physical hardware; other traffic participants target rival cars, whether they want to set up a few cars or as many as dozens or even hundreds of cars, Digital twins can be used to build the digital models of these vehicles into the digital twin environment, including the performance of the on-board V2X communication module, and their models can also be built through PreScan digital twins, or through technologies such as radio frequency simulation. To do a test that combines virtual and real, through these technical means, the efficiency of the test can be improved and the cost of the test can be reduced.
Second, through the combination of virtual and real, some dangerous scenes can also be simulated, which can only be encountered by chance on some roads, but the scenes that are difficult to appear in the experimental field.
Third, there are some basic scenarios, but when you want to carry out high-coverage divergent tests, you can also use digital twins. Build the behavior and distribution of any traffic participants we want around the host car to improve test coverage.
Q7. Compared with other simulation software companies of the same type, what are the advantages of Siemens?
Huang Hanzhi: I will talk about my understanding. I also noticed that there are many companies in the market recently that have entered the field of digital simulation of autonomous driving. Compared with these “rising stars”, Siemens’ PreScan is relatively mature in terms of brand and technology, and it is also more systematic. The digital twin platform for autonomous driving with full links can fully satisfy customers’ needs for virtual simulation, The need for physical testing and verification means has now been the most widely recognized and applied.
Q8. What layout does Siemens have in establishing and cooperating with local scene libraries in China?
Huang Hanzhi: Whether in the world or in China, there are some excellent projects that are studying the definition logic and data standards of the scene library. For example, the ASAM organization (Siemens and Siemens Digital Industry Software are also members of the technical committee) lead the definition of some international common Open series scene library file data formats.
We are doing in-depth and comprehensive cooperation with international and local scene database data companies. Whether it is at the level of technical access or at the level of commercial customer projects, there are many successful examples.
Q9. What localized development has Siemens made for China’s traffic environment?
Huang Hanzhi: Let me give two small examples.
the first
One, in Siemens’ traffic scene model library, there are many traffic participants with Chinese characteristics. For example, in the release of the official version of PreScan, there are digital models of SAIC-GM-Wuling’s very popular models, as well as Nanjing Jinlong Kaiwo. Models of unmanned shuttle vehicles, as well as models of earthmoving vehicles and dump trucks unique to China.
Second, in addition to the above elements in our scene, we also need to build different behaviors. For local traffic participants in China, Siemens has done a lot of localized research and development on the distribution of driving behavior patterns on the road.
Q10. In terms of sensor simulation, what is the recent progress of PreScan? What features are you going to support next? For 4D millimeter wave radar, is there support for various route lidars?
Huang Hanzhi: PreScan has always been an absolute technological leader in the field of environmental perception sensors. Whether it is a vehicle-mounted camera, or millimeter-wave radar, ultrasonic, lidar, and V2X fields, a very in-depth model of physical perception characteristics has been established.
There is support for 4D millimeter wave radar and various lidars. Regarding millimeter-wave radar, there are very complete characteristic models in terms of target-level simulation, target perception accuracy, resolution, energy-level simulation, and original signal-level simulation.
On the lidar model, the infrared band that determines the perception performance of lidar, beam energy and beam size, scanning mode, beam propagation attenuation in different media, or reflectivity under different materials, shapes, colors, and target surface characteristics , as well as the signal receiving efficiency of lidar and other characteristics are supported. Lidars for various technology paths are supported.
Q11. Please look forward to the future development trend in the field of intelligent networked vehicle simulation.
Huang Hanzhi: In the field of autonomous driving simulation, I think we are a well-deserved leader.
Siemens is developing very fast in many fields such as digital R&D, digital manufacturing, digital management, and digital services. In the field of automobiles, whether it is digital simulation of autonomous driving, software-defined vehicles, and software life cycle management, they are also developing rapidly.
In terms of development trends, first of all, the simulation of intelligent networked vehicles requires not only simulation development tools for the intelligent networked R&D department, but also with the digital R&D platform of the enterprise, the data management platform of the enterprise, and the software life cycle management platform of the enterprise. This is the first development trend in order to maximize the digital advantages of ICV simulation.
The second development trend, the automotive autonomous driving industry is looking forward to a mature, stable, fully functional, open platform, digital simulation application environment that supports cloud computing platforms, all of which are our development direction.
Regarding the Prescan software tool itself, we have several development trends. The first is a more physical and accurate digital twin model; the second is a more flexible and efficient cloud computing platform; the third is to meet functional safety and The creation and search of massive scenarios required by SOTIF, as well as the extraction of key scenarios, focus on development; finally, build a digital R&D tool chain at the enterprise development platform level to improve enterprise R&D efficiency, reduce R&D costs, and achieve optimal autonomous driving system performance. Enterprises capable of autonomous driving can face the challenges of digitalization and become successful enterprises under the wave of digitalization.
