Using AI to control the chassis, what big game is Huawei playing?

"The control-by-wire chassis is an inevitable requirement for advanced autonomous driving, and will undoubtedly be the dominant direction in the development of future automotive chassis systems."

Zhou Jianguang, chief technology officer of China Automotive Chuangzhi, unveiled the importance of intelligent chassis in one sentence at the 2023 China Automobile Supply Chain Conference. Control-by-wire chassis, as the only way for the development of automobile intelligent chassis, is an electronic transformation of traditional automobile chassis. After electrical signals replace the mechanical signals of the traditional chassis, the control of the car can achieve "human-machine decoupling."

At present, the chassis is already developing from electronic to intelligent. Under the precise control of automobile sensors, intelligent control clusters, and computing chips on chassis configurations such as automobile suspensions and motors, intelligent driving control supported by intelligent chassis is the third major technology for smart cars after intelligent driving and intelligent cockpit. plate.

"Intelligent chassis is the sector that can really directly change the driving experience." Liu Diqing, product director of Zhijie S7, believes.

Take the Huawei Touring intelligent chassis (hereinafter referred to as Touring chassis) installed on the Zhijie S7 as an example. Its CDC (variable damping shock absorber) can sense the road surface 1,000 times per second and adjust the damping 100 times, which can reduce the maximum Nearly 60% impact. Its air suspension can automatically adjust the height and posture of the vehicle body according to vehicle speed and different driving conditions.

In this way, driving a car is like driving a "humane old horse". The driver only needs to control the speed and direction, and the car can handle other details such as bumps, obstacle avoidance, and cornering by itself.

Yu Chengdong introduced the intelligent chassis at   the Zhijie S7 ( parameters | inquiry ) press conference. In the era of software-defined cars, the Touring chassis amplifies the advantages of intelligent sensing and intelligent control that Huawei is already good at, allowing smart cars to provide fuel and fuel. The car’s different handling, comfort, braking, safety, etc. driving experiences.

In 2023, when automobile intelligence is officially recognized by consumers, the experience provided by smart chassis will further amplify the cost-effectiveness of smart cars. On the other side, players who also smell the new trend of smart chassis are also stepping up their layout. For example, BYD launched Yunnan earlier, and CATL will soon launch a smart skateboard chassis. Xiaomi Group, BAIC Industrial Investment, NIO Capital, etc. will also invest in a number of automotive smart chassis-related companies in 2023.

The bets made by automobile giants have transformed the intelligent chassis of automobiles into a new battleground for intelligent automobiles after smart driving and smart warehouses.

No bumps or shakes, a million-level driving experience for 300,000 yuan?

I still remember that afternoon twenty years ago when I was riding a tour bus through a dirt road full of sharp turns and potholes. After the tour guide asked everyone to endure the "special massage road" for a few minutes, almost all the passengers on the bus began to get motion sickness.

It would be great if the car could be ridden without bumps or shaking.

In order to achieve the goal of "no bumps and no shaking", car companies in the past usually worked hard on the chassis. For example, Mercedes-Benz uses air suspension + adjustable damping shock absorbers on the Mercedes-Benz E-Class to enhance the shock absorption effect. The Mercedes-Benz S-Class is also equipped with an active suspension system that can detect road conditions ahead and adjust the "softness and hardness" of the suspension in advance, ultimately achieving the effect of reducing longitudinal vibrations of the car. In order to reduce the lateral "sway" perception, BMW introduced torque vectoring on the X5 and X6, allowing the car's steering to distribute power more reasonably based on the differential, thereby achieving a more stable turning effect.

The solutions of Mercedes-Benz and BMW were the leading ones in the past era of fuel vehicles. But now, if we look at the past kings, their products often cost hundreds of thousands or even millions more. The sensors rely more on passive sensing adaptation strategies, which are somewhat unsuitable for the current involution of China's smart car market.

Nowadays, the car riding experience has achieved the same effect as that of Mercedes-Benz and BMW. Huawei has priced the product on the Zhijie S7 to less than 300,000 yuan. This is due to the ability of "software-defined chassis" - the multi-modal perception and multi-domain collaborative control of the Touring chassis.

The first is the overall hardware configuration. Huawei has equipped the Touring chassis with a DriveONE 800V silicon carbide high-voltage power platform. It adopts a front double wishbone and rear five-link independent suspension design, with lateral stabilizer bars and CDC variable damping shock absorbers. and air suspension.

Dismantling the hardware components of the Touring chassis, the power platform mainly consists of the motor that provides power and the efficient charging and discharging platform. The front double-wishbone rear five-link independent suspension mainly uses the front virtual kingpin double-wishbone's effects of less roll, less impact, and high stability to improve the chassis' anti-disturbance capabilities. Combined with the rear five-link independent suspension, it takes into account both comfort and control.

In order to reduce the longitudinal vibration of the car, the Touring chassis has built the ability of CDC+ air suspension to adapt to the car's working conditions. With the sensor connected to achieve up to 1,000 road surface sensing per second, the CDC control valve can perform fine damping adjustment control up to a hundred times per second based on the road surface sensing results. At the same time, with the cooperation of CDC, the air suspension can also automatically adjust the height and attitude of the car body. Taken together, the CDC shock absorber can reduce the impact force by up to about 60%.

On the other side, in order to reduce the lateral "sway" of the car, the Touring chassis has optimized the previous TVC (Torque Vector Control) solution in the industry. The principle of TVC is that after the car autonomously scans the road surface 100 times per second, it will automatically determine whether the car will be underpowered when cornering. Then, based on the calculation results, the system will redistribute the engine power to the drive wheels by controlling the differential gear, and continuously adjust the specific power output based on the actual observed values ​​during the cornering process, thereby achieving smoother cornering and fewer passengers. Body shaking.

After the optimization of Touring's chassis, the original TVC's "three-step process" of "estimate calculation, power distribution, and adjustment while running" has added more accurate judgment capabilities.

As can be seen from the control chart of the "Sports +" mode, the Touring chassis display has added a curve detection module to identify more accurate geometric features of the current curve in real time. Use more accurate curve modeling to cope with actual road conditions that are far more complex in shape than the test site. On the other side, the Touring chassis distinguishes the car's entry into a corner and its exit from a corner, and controls the power distribution more carefully to achieve push-in or tail-out of a corner.

In other words, in the TVC after the Touring chassis is optimized, the car's lateral sway control can be more refined. In this way, the performance of the car can be improved by 2-7 percentage points under the conditions of the test site. The car ride experience (horizontal and longitudinal acceleration) is better than Taycan 4S and Model S.

Judging from the final implementation effect, after combining and calculating more and more complex sensor data, the Touring chassis has achieved an optimized ride experience of "no bumps and no shaking" of the car.

However, the intelligent transformation of the Touring chassis compared to the traditional car chassis is not limited to the improvement of sensors and on-board computing power. The multi-modal perception and control capabilities from the comprehensive automotive smart warehouse, smart driving, and smart control are the real "power" of the Touring chassis.

To define an intelligent chassis, hardware is the foundation and AI is the upper limit.

"Algorithms are the most important." Liu Diqing emphasized.

Just like a good mobile phone system, it can reasonably perceive and control the ability to deploy software and hardware, and maximize the control experience of the mobile phone.

Smart cars also require the perception and fine control capabilities of smart chassis to combine and maximize car hardware, power, sensors, smart driving and other functions.

First, at the level of car perception, Touring Chassis introduces MFSS (Multi-modal Fusion Sensing System) to answer the questions of "who am I" and "what do I want to do" for the car.

In vehicle state sensing (iVSE), the principle is to first obtain an estimate of the current state of the car through the interaction of multiple sets of internal sensors on the vehicle. Next, the car combines the road information collected by external sensors to completely determine the current "place" of the car, and finally controls the car.

On the other hand, in car road surface preview (RSS), the car identifies road surface information in advance through the camera (or lidar) of the intelligent driving perception system. Thus, the CDC can be adjusted in advance according to the different conditions of the road surface (road surface material, friction, roughness, etc.). Therefore, it can achieve "predictive control" compared with the previous passive detection solution.