On December 30th, at the Great Wall Motors’ online Intelligent Driving Strategy Upgrade conference, Zhang Kai shared the following speech:
Hello everyone, I’m Zhang Kai from Great Wall Motors. Today, I am honored to be here and share with you the innovations and progress of Great Wall Motors in the field of intelligent driving. At the same time, I would like to thank you all again for taking the time to participate in our conference.
As we know, Café Intelligence is the intelligent brand that Great Wall Motors released in July this year. It covers intelligent cockpit, intelligent driving, and intelligent electronic architecture. We also proposed the R&D concept of “Three Intelligences Fusion” for the first time in the industry.
Today, we are proud to announce Great Wall Motors’ new proposition in the field of intelligent driving — Café Intelligent Driving. Please take a look at the big screen.
We will take three years to achieve three leading positions: the largest user scale in the industry, the highest user experience rating, and the most comprehensive scenario coverage, in order to support our position as the leader in autonomous driving in the intelligent era in China. This is the Café Intelligent Driving 331 strategy that we are announcing today.
Café Intelligent Driving’s “331 Strategy” is based on Great Wall Motors’ deep insight into user needs and active thinking about providing unlimited possibilities for users in the future, which is the belief of Café Intelligent Driving.
The “331 Strategy” of Café Intelligent Driving is based on the accumulation and development of Great Wall Motors in the field of intelligent driving over the past ten years, as well as the active exploration and pursuit of future intelligent technology. This is the confidence of Café Intelligent Driving to become the leader of autonomous driving in the intelligent era.
Ten years of accumulation and development have created a full-stack self-developed technology in the field of intelligent driving. As you can see, not only have we self-developed the automatic driving software system, but we have also adopted self-developed mode for the automatic driving controller and middleware system.
It can be said that we have done everything except the OS and chip, because we believe that the current automotive industry is rapidly becoming internet-based. A significant characteristic of the Internet industry is the speed of iteration, and what we pursue is to be fast.
In terms of underlying hardware, we have developed a series of self-developed autonomous driving computing platforms to match our different price-range models. We have an autonomous driving computing platform with ultra-high computing power, using Qualcomm chips to build the industry’s highest computing platform. At the same time, we also have an autonomous driving computing platform that considers cost and computing power balance, enabling users who purchase our lower-priced models to enjoy the fun of autonomous driving functions.
In terms of middleware, we began to develop middleware systems based on AP software in early 2018. It can be said that we are the earliest domestic automakers to conduct research and development on AP software.
Middleware is a very important part of autonomous driving software, which directly affects the efficiency and effectiveness of autonomous driving software running on the controller. The middleware system can perform necessary monitoring and real-time diagnostics of the autonomous driving software, and can also better leverage the computing power of the domain controller chip.
In terms of autonomous driving software, we have completed not only traditional ADAS functions such as ACC/AEB, but also engineering deployment of advanced autonomous driving perception, fusion, prediction, planning, and decision-making software in the high-level autonomous driving field, and have completed millions of kilometers of road testing.
In terms of closed-loop data, another significant feature of the Internet industry is the importance placed on data. By mining data, we can quickly find the corresponding scenarios we need; by mining data, we can quickly optimize the performance parameters of algorithms, making the performance of autonomous driving functions more in line with user habits. Currently, all of our computing platforms have realized real-time processing and feedback of data.
We believe that the development of autonomous driving systems depends on accumulation. Data needs to be accumulated, scenario capabilities need to be accumulated, and stable and reliable functions require continuous accumulation. Below, I will take the example of the implementation of the AEB function to illustrate the importance of knowledge accumulation in the field of intelligent driving.Previously, after communicating with many industry experts, it was commonly believed that Automatic Emergency Braking (AEB) is one of the most difficult functions to implement in autonomous driving software. The media once described it as “After all this talk about autonomous driving, why is it still so difficult to achieve the AEB function for safety?” In fact, we also believe that the AEB function is one of the most challenging functions to balance in the implementation of autonomous driving functions.
The AEB function is based on sensors located at the front of the vehicle which detect possible collision risks with other vehicles, pedestrians, or other traffic participants. It automatically triggers the vehicle’s execution mechanism to apply the brakes and avoid the collision.
We can see that the AEB function was officially incorporated into the E-NCAP evaluation regulations in 2014. By 2016, the evaluation regulations introduced protection for road users, also known as AEB pedestrian protection. By 2018, the evaluation regulations introduced protection for cyclists. Until now, the evaluation regulations continue to incorporate AEB functions for more complex scenarios, such as AEB-intersections, AEB-turning, AEB-pedestrian avoidance, and AEB-reversing.
As time passes, regulations continue to progress and we face increasing technological difficulties while also demanding higher performance from vehicle sensors.
We believe the complexity of AEB system development lies in three aspects. Firstly, the AEB system has the shortest operating cycle of only 20ms, while other autonomous driving systems commonly have operating cycles of 50-100ms. Completing target identification, vehicle motion decision-making, and planning within 20ms is already a very difficult task.
Secondly, the AEB system has extremely strict requirements on target recognition, with over 30 requirements for each target attribute. The system has different calculation logics for each target type, motion direction, and motion state. For example, the calculation logic for longitudinal traveling vehicles and pedestrians crossing the road is very different, so it is necessary to ensure that each scene can be correctly triggered. Additionally, ensuring the correctness of arbitration execution when multiple scenes are triggered simultaneously is also a very difficult task.
Lastly, the control strategy of the AEB system is highly correlated with the hardware execution. During the execution process of the AEB system, the braking force of the entire vehicle is very strong. Once there is a mistaken trigger, it can cause great panic for the driver and passengers. Therefore, the design logic of the target selection, risk assessment, decision-making control, and other modules of the AEB system must be rigorous.The control strategy of the AEB system must be highly coupled with the braking capability of the hardware actuator to achieve the desired design effect. And all of this requires years of accumulation. Years of technological accumulation have made our autonomous driving technology safer.
With years of technological accumulation, deep insight into user needs, and active exploration and pursuit of future technologies, Coffee Intelligence Driving has formed a relatively complete technical system. Next, Coffee Intelligence Driving will be prioritized in WEY flagship models, and the new car will be released in early 2021.
In 2021, the WEY brand of Great Wall Motors will be fully equipped with Coffee Intelligence Driving. Starting in the second half of the same year, other brand models of Great Wall Motors will also be equipped with Coffee Intelligence Driving one after another.
Our ultimate goal is to allow drivers to enjoy a cup of coffee with peace of mind in a self-driving car. We believe that Coffee Intelligence Driving will become a powerful support for the development of Great Wall Motors into a global technology travel company, and a strong booster for the development of Great Wall Motors towards the future, aiming to win the world!
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.