From March 26 to 28, 2024, the 2024 China Commercial Vehicle Forum was held in Shiyan City, Hubei Province. Hosted by the China Association of Automobile Manufacturers, with the theme of "New Pace, New Results, New Improvement, Helping the High-quality Development of the Commercial Vehicle Industry", this forum is based on the high-quality development requirements of the industry, the realization of the national "double carbon" goal, the transformation and innovation needs of the automobile industry, and promotes reform, transformation and development with innovation, so as to help the high-quality development of the commercial vehicle industry. Among them, on the morning of March 28, Yang Bo, technical director of Suzhou Zhixing Zhongwei Intelligent Technology Co., Ltd., delivered a wonderful speech at the "Theme Forum II: Intelligent Network Technology Empowerment and Accelerating the Development of the Commercial Vehicle Industry". The following is a transcript of the speech:
Distinguished guests, leaders, good morning.
My name is Yang Bo, I am from Suzhou Zhixing Zhongwei Intelligent Technology Co., Ltd., and I am very happy to share with you some of the work and achievements of IAE in the vehicle in the environmental system in the past few years.
Today, I will share through three aspects, first, I will first introduce our company's IAE, including our X-in-loop technology system and solutions. In addition, we will introduce a system of the whole vehicle-in-the-loop. Finally, I would like to report to you about the application and practice of our industry as a whole in the past few years.
First of all, our company headquarters is located in Suzhou, currently in major cities in China, including in Europe have relevant R&D centers, personnel, we are a supplier focusing on the entire field of intelligent driving simulation testing, our goal is to start from the perspective of the commonality of the entire industry, and then to create and improve our simulation test from a large number of scenarios to the test of extreme scenarios, forming a closed loop of technology, a closed loop of data, better to help, to empower the commercialization of our entire intelligent driving safety.
In order to enable our intelligent driving to land safely, our vision, I believe that all of you here, we have a common vision, which is to build the confidence of the entire intelligent driving safety, including credibility. In fact, it is very easy to understand that we need to spend a lot of money on the road through the way we use real cars in the field and public road tests, whether it is the cost of money or the cost of time. Therefore, we expect that through this scientific and effective simulation test technology, the test and verification in the field of intelligent driving can be shortened as much as possible, and the cost of the industry can be reduced as much as possible. Therefore, we are now building a well-architected scenario and database resources, which can be simulated through our cloud platform, and then we can do our cloud-based massive scenario simulation test, vehicle-in-the-loop system verification, and finally based on standards and regulations to carry out the final vehicle site evaluation and certification.
It is my honor to report to you that in the past few years, we have built such a complete X-in-loop technology closed loop, including a data closed loop system. First of all, we can import the data collected by the field, that is, the vehicle side, including the data that we can now collect based on the road test unit, and then import it into the virtual simulation platform for generating de-simulation test scenarios, which currently correspond to our company's Shuimu Lingjing scene factory. On the one hand, the scenario on the simulation side supports the training of perception algorithms and generates relevant training datasets. On the other hand, we can support these scenarios for the entire intelligent driving field, we say that in the automotive V-shaped development process, including ADAS SIL/MIL, driver-in-the-loop DIL, and the entire vehicle-in-the-loop, these scenarios can be used in the development process verification system. Finally, we can also support companies to do site testing, including road testing. Therefore, with such an X-in-loop technology system, it also means that we can achieve a closed loop of data, including a closed loop of technology, in the field of simulation and testing.
Second, the vehicle-in-the-loop technology system.
In fact, our intelligent driving, especially commercial vehicles before landing, needs to do a lot of this kind of testing and verification, and the video of the actual field test shown by the guests just now is also quite exciting. In terms of our IAE philosophy, in order to reduce costs and shorten the time cycle, we actually hope to use the new three-pillar method to support such a system for simulation testing and evaluation of autonomous driving. The so-called new three-pillar law consists of three links. First of all, there is a massive simulation platform with cloud computing power to support us to do high-mileage simulation tests for massive scenarios, which can improve the coverage of our overall testing, including the fact that we can extract the so-called dangerous scenarios in the results of cloud-based simulation tests.
Next, we will use advanced vehicle-in-the-loop solutions, including our overall vehicle-in-the-loop system, and we will verify its decision-making strategy, control, and actual response to the extreme scenarios and dangerous scenarios that have just been screened out, all of which can be realized on the ring bench.
We are divided into three types of vehicle-in-the-loop technology systems, first of all, the first category of PG-VIL, site-based vehicle-in-the-loop technology. In the middle is VaHIL, a high-end vehicle-in-the-loop technology solution based on the laboratory. The last VTHILS traffic-environment-vehicle-in-the-loop, large-scale laboratory program. Why the entire vehicle is divided into these three types is closely related to our entire intelligent driving system.
In fact, an analogy with a human driver, an intelligent driving system, is nothing more than divided into our front-end perception system, followed by our brain, decision-making layer, and the last is our actual execution layer, to perform the actual steering, acceleration and deceleration of these actions. When so many factors are coupled, it is actually difficult for us to analyze the problem at the level of the whole vehicle. Therefore, in our entire vehicle-in-the-loop technology system, we actually need to decouple these aspects just mentioned. Our different test solutions are oriented to different test focuses, they are irreplaceable, they are complementary to each other, for example, our VTHILS large laboratory, we test for environment-related, sensor-related, including some expected functional safety scenarios. Our VaHIL is mainly for the decision-making level, including some high-speed, dangerous scenarios into the laboratory, which will ensure that the relevant tests are done in a safe situation. Finally, we have PG-VIL an execution-oriented, control-responsive system in this area.
After having the entire vehicle-in-the-loop technical system, which scenarios we want to measure, this is our very important data support. In fact, many experts have also mentioned that we will disassemble some scenarios that we are more concerned about for the functional specifications of the entire commercial vehicle. Our company began to do such accumulation work related to the scene a few years ago, and the table you see now, as of the beginning of this year, we have accumulated the scene resources, then this scene resources include our so-called digital twin 1:1 in the simulation software side of the high-precision scene, currently more than 1,000 kilometers. And then there are some regulations, including V2X scenarios and so on, and we have very rich scenario resources. And our scene is still in production, so in fact, the number of scenes we have here is continuously updated on a daily basis.
Next, I will show you here, which is the video of the scene we have now, that is, as some experts have just mentioned, now commercial vehicles are used in this kind of port, this kind of application in mining areas, and then include commercial vehicles on the highway to pilot these functional scenarios, including we also have some test resources for agricultural machinery scenarios. As I just mentioned, you can see that there are so many scenarios in the table, and then we are talking about how to have a very efficient tool to support us to do this policy, so this piece mentions the data of our new three-pillar method, or the support of the tool, which is the simulation platform of our jellyfish cloud. In this part, our company has cooperated with Alibaba Cloud and China Mobile to build a SaaS platform to serve the entire industry based on massive simulation cloud testing based on cloud computing power. Assuming that 400 cloud-based nodes are used as an example, such a simulation test effect of millions of kilometers per day can be achieved. This is an introduction video about our jellyfish cloud simulation platform, you can take a look, in fact, it is an integrated virtual simulation platform testing tool, we will have tens of thousands of basic scenarios, and then we have the corresponding web side to support you to upload vehicle dynamics models, sensor models, so as to do iterative testing of algorithms, and then include the module we have the evaluation can output these parameters of the entire intelligent driving system we are concerned about with one click.
Third, the application and practice of vehicle-in-the-environment technology industry.
First of all, I will briefly describe to you the system principle of our different vehicle-in-the-loop technology, first of all, about the entire PG-VIL system, we have a real vehicle running in a real proving ground, the target is not a dummy car system, through the simulation environment to build a virtual system, the sensor through the injection form, back to the vehicle controller. A real vehicle in an open field, after activating the intelligent driving function, if there is a pedestrian suddenly in front of it, at this time the real vehicle can achieve the effect of braking or steering in this open field, the reason why this piece puts it, the whole target through the form of injection, mainly in several considerations, the first is safety, so that it will not be for us, than even if it is a moving target, and then we will not involve the risk of self-car. The other part is that the whole test is very repeatable, and it is very convenient to implement this complex and rich test cases through the form of simulation injection.
From the level of system architecture, we will involve some positioning-related development, including these injections of sensors, we are through bus injection technology, so that the signal back to the controller of the vehicle. As you can see, the sensors used for intelligent driving in the table on the right can all be used to inject relevant signals.
From the perspective of typical application, you can imagine that if we have developed a set of intelligent driving system, we now need to care about its actual implementation, in fact, if we do the test at the actual site, this time is often very limited, or very inconvenient, because we need to go with our main car and the target, mobile platform, in advance to define and design the interaction logic. But it is very convenient for us to interact with each other in the form of triggers. On the other hand, for example, we are now facing the AEB system, especially for our commercial vehicles, in the end in this AEB system, whether the control instructions output by the algorithm can be executed by the whole system, whether the calibration parameters can meet the needs of the whole facing after we have a moving target, the effect of successful braking is actually unknown, we realize the verification of the response of the real vehicle through the technical scheme of PG-VIL, including the verification of the rationality of the entire calibration parameters. So in general, the technical solution of PG-VIL is oriented to our decision-making, control response, including test and verification related to system calibration.
Next up is our VaHIL, the system principle for advanced vehicle-in-the-loop. In fact, we divide it into L2 and L3 according to the level of the entire intelligent driving. The main difference is that in L2, we need to provide relevant visual input to the driver, and provide a virtual picture of what the driver sees, so we will add a visual simulation system to this piece. The rest of the virtual sensors, these are built in the simulation environment, through the sensor simulation system and then back to the vehicle controller. This is a system architecture, and we can see that the real vehicle is actually fixed to the bench, which is relatively forbidden to the ground. And then we have a simulation system that creates the simulation test scenarios that we care about now. Then these sensor-related signals used to deactivate the intelligent driving function can be injected back to the vehicle controller in the form of injection or combined with the form of sensor simulator, so this is our advanced vehicle-in-the-loop system architecture.
From the perspective of typical application, you can imagine with me again, now there is a commercial vehicle to do a high-speed driving related function test, this test can be done on the actual road in the field, but often the first is the safety factor, the other is an efficiency problem. What kind of advantages do we have when we move it in-house? First of all, I use simulation to create this kind of highway mix, and the complex traffic flow is actually very convenient, and we can use this trigger form to define the interaction logic between the trigger and the target vehicle. In addition, the whole vehicle is relatively stationary, and the high-speed operation is not limited by any site, so I don't need to prepare a 50-kilometer, 100-kilometer site specifically for this. On the other hand, in the early stage of R&D, in fact, our system still has security risks, at this time, when we have functional failures, downgrades, including the withdrawal of functions, there is actually no security risk to people at this level.
So on the whole, our VaHIL solution uses typical high speeds, including extreme operating conditions, including we can also do some tests like sudden brake failure.
Finally, I would like to introduce VTHILS, a real laboratory environment to create a real environment, rain, things, light, ground water, icing and other environmental effects, so as to achieve a real and controllable environment, we can help intelligent networked vehicles in such an environment to verify the ability to work safely in bad weather, the site requirements of this laboratory will be relatively high, generally speaking, for commercial vehicles require such a size of more than 700 meters. From the perspective of the whole system architecture, the most important thing is the environmental simulation system, which is mainly divided into rainfall simulation, precipitation simulation, water cycle, snowfall, lighting and so on. After we have these simulation systems, we also need to implement closed-loop monitoring of the simulation effect, including control, and we have the environment including video monitoring, as well as ancillary facilities, because the exhaust air, electricity, and signal lights inside the laboratory need to be managed. In general, for these three large systems, there is a central control system to centralize the dispatch to handle these related controls, including monitoring functions.
Therefore, from the typical application of VTHILS, we can do the test and verification of the performance limitations of the sensor in the environment and weather controllable environment, including the test of combining the environment and regulations, because now many tests are actually in the ideal environment, sunny days, daytime, and road conditions are also very good. For our intelligent driving system, such as what is the solution in the case of dense fog, this also supports our testing. It is a detection effect of the detection rate and recognition rate of the target by the perception algorithm in the range of vehicle speed under a dense fog condition.
That's just a bit of an introduction to our overall IAE, X-in-loop, including our vehicle-in-the-loop technology system. In fact, having said all this, how have we empowered the entire industry in the past few years? In fact, I can share with you, that is, some videos of some of our actual projects in this area. Including our use of PG-VIL verification ACC, AEB. Then we do some high-speed test and verification of the above functions on the whole VaHIL system, including our verification of VTHILS, sensor performance, and limitations. I am very happy to have completed these things with my industry colleagues and some of our customers, and I have also landed a lot of projects, and then I really feel that I am promoting the development of our entire intelligent connected industry. Therefore, I hope that I will have the opportunity to cooperate with all of you here, and then promote the accelerated development of our entire intelligent network.
That's my debriefing today, thank you.
(Note: This article is based on on-site shorthand and has not been reviewed by the speaker)
Source: China Commercial Vehicle Forum
![Learn to learn from Jackie Chan like a man! I have a girlfriend who kisses others, and Yang Bo is a talk show: She forced me to break my shoes[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)