As intelligence and electrification have become the main themes in the Chinese auto market, Mercedes-Benz, which has been deeply rooted in the Chinese market for 38 years, has demonstrated the "China speed" of the new era.
In April last year, No. 399 Keqiao Road, Pudong New District, Shanghai was still a building with an unfinished facade. Just 12 months later, the new building of Mercedes-Benz Shanghai R&D Center, which was located here, opened.
With the new building in operation, Mercedes-Benz's R&D layout in China has officially entered the third phase - using Chinese innovation to lead the global trend. In simple terms, it is to apply the R&D results of the Chinese team to the global market.
If the Mercedes-Benz Beijing R&D Center in Yizhuang, Beijing is mainly responsible for checking the mechanical quality of a car, then the Mercedes-Benz Shanghai R&D Center is the "strongest brain" of Mercedes-Benz in the era of smart electric vehicles. The core technologies in the electric field such as intelligent driving, intelligent cabin, and three-electric are all developed here.
Mercedes-Benz recently held a Technology Day event at its Shanghai R&D Center, showcasing its latest technology that will soon be installed in mass-produced vehicles. The core of these technologies is Mercedes-Benz's self-developed MB.OS architecture and the "mapless" L2++ autonomous driving system, both of which will be installed in Mercedes-Benz's new cars to be launched next year.
In the field of autonomous driving, the current trend has shifted from traditional regulatory decision-making to end-to-end large models. The biggest difference between the two is that the former is based on code and performs specific actions according to specific scenarios, while the latter has the ability to learn and make adaptable actions based on actual road conditions.
According to reports, the "mapless" L2++ autonomous driving system will use a new generation of sensor hardware to significantly increase the upper limit of perception capabilities. The algorithm has evolved from BEV+transformer+occupancy network to an end-to-end large model. In addition, all perception training data of the large model uses Chinese data and adds a large amount of training materials from Chinese veteran drivers, so it can master the differences in traffic infrastructure such as traffic lights, road signs, and road markings in different regions of China.
Inside the R&D center, The Paper reporters noticed that the intelligent driving test bench used by Mercedes-Benz can simulate the most extreme road conditions through digital twin technology, "educating" the system to be more sophisticated in a more efficient way.
Of course, in addition to simulated road condition tests, Mercedes-Benz also conducts a large number of actual road tests to ensure the performance of the system in a real environment.
The road test video played by Mercedes-Benz at the Technology Day showed that the self-driving test car using an end-to-end large model was very experienced in making unprotected turns at intersections. For example, when turning right, there were non-motor vehicles constantly shuttling sideways. The test car did not wait in place as before, but played a game with these non-motor vehicles and slowly moved forward until the turn was completed. This behavior pattern is very close to that of an "experienced driver".
It is worth mentioning that Mercedes-Benz recently announced that it has become the first international automaker to be approved to conduct L4 urban and highway autonomous driving tests in Beijing. Previously, it was also the first international automaker to obtain an L3 road test license.
In addition to driving, the new generation of autonomous driving system has also made a qualitative leap in parking. It also uses Chinese parking training materials, can calmly deal with China's complex parking scenarios (dead-end roads, three-sided walls, etc.), small-sized parking spaces, can achieve line parking, and can identify special-shaped obstacles such as ground locks and A-shaped signs.
In the demonstration video, the huge Mercedes-Benz EQS flexibly maneuvers in a narrow parking space, demonstrating a parking skill level that is more efficient than that of most humans. More importantly, the system no longer requires mandatory parking directions, and both the front and rear of the car can be parked in and out.
Another major technological achievement is the MB.OS operating system. Strictly speaking, it is not just an operating system, but a new architecture that covers the vehicle end to the cloud, realizes the decoupling of software and hardware, and covers intelligent driving/intelligent cabin/body comfort and communication modules. It is also the "central nervous system" of future Mercedes-Benz cars.
This architecture is first of all the industry leader in hardware. For example, the smart cockpit domain uses Qualcomm 8295 main chip, which has a 72% increase in CPU computing power and a 91% increase in GPU computing power compared to the previous generation in the same package space; the autonomous driving domain uses NVIDIA DRIVE Orin chip, with a single computing power of 254TOPS and a system safety level of ASIL-D.
Mercedes-Benz leads the system design and construction in terms of software, with the core strategy of safely controlling and processing vehicle and cloud data. At the same time, MB.OS's functional applications, OTA, safety functions, etc. will carry more complex system functions with more advanced communication capabilities.
Specifically at the application layer, the new generation of voice systems supported by the MB.OS architecture have become more agile, flexible and user-friendly.
In terms of parameters, the response speed of the new generation of voice system has increased by 1.5 times, and the response time is only 0.2 seconds; the new system can accurately identify the instructions of people in each seat, and can be interrupted and continued at any time; in terms of the "voice" of AI, the new system also adopts more human and realistic replies to enhance the interactive experience.
It is understood that the training data volume of the new generation of voice system has increased by 2-3 times, and it can better understand verbal instructions. For example, it used to be necessary to wake up the system by saying "Hello Mercedes-Benz" and then issue the command "Turn on the air conditioner", but now you can say "Turn on the air conditioner to 24 degrees, Mercedes-Benz".
Based on the MB.OS architecture, the UI interface in Mercedes-Benz's future new cars will also undergo major changes. In the demonstration video watched on site, Mercedes-Benz did not use the "flat" icons that are popular in current tablets or mobile devices, but still used three-dimensional icons with rich light and shadow refraction to express a sense of luxury. At the same time, the overall interface uses translucent materials, which are visually light and transparent, and effectively reduce visual pressure.
On the navigation system interface, the new system deeply integrates navigation information with the intelligent driving SR interface, and displays important information directly at the driver's best observation position to ensure efficient information acquisition. In the demonstration video, the new navigation interface is more in line with the current habits of Chinese people. The system is divided into intelligent driving state and driver driving state. In the human driving scene, more emphasis is placed on navigation elements that support human driving; in the intelligent driving scene, more emphasis is placed on the perception and decision-making information of intelligent driving, and the appropriate display field of view can be automatically adjusted according to the scene.
Objectively speaking, these intelligent technologies, which will be launched together with the new cars on the MMA platform next year, will allow Mercedes-Benz to catch up with the level of the first-tier new forces in terms of "coolness", but it is difficult to say that they will lead the market.
However, the dimensions for evaluating whether a car is good are not all about intelligence. The basic three-electric system and quality are where the real skills are demonstrated.
Although the research and development stage can be accelerated, traditional luxury brands including Mercedes-Benz will resolutely refuse to skip the most important verification step.
Taking the salt spray test to verify the durability of batteries in harsh environments as an example, Mercedes-Benz has added a high and low temperature cycle test (-40°C to +50°C) to simulate the real extreme application scenarios encountered by customers. The test will last for 8 weeks without interruption for a second, a total of more than 1,300 hours of corrosion testing, which is 9 times the national standard requirement of "144 hours".
In terms of battery aging testing, Mercedes-Benz has to conduct at least "two winters and two summers" of testing, and a single test takes two and a half months.
According to a recent report released by consulting firm AlixPartners, new forces launch products faster than traditional car companies, mainly because of the difference in the verification process. The verification period for the former is about 6 months, while the latter is as long as 18 months. This 12-month gap cannot be felt when using the car for the first time, but it will always leave traces in the long river of time.
As the inventor of the automobile, Mercedes-Benz knows well which areas can be accelerated and which areas need to adhere to the long-termism of "slow is fast".
