As high-end intelligent driving gradually sinks from high-end models to the low- and mid-priced market, the target market for core sensors, especially lidar and 4D imaging millimeter-wave radar, is also seeking stratification and integration.
"Lidar only emerged after the emergence of autonomous driving. In the early stages of the development of these systems, many companies spared no effort to make the sensors as powerful as possible. But entering the second half of the competition, the focus has shifted to creating tangible value." Founder of Aotu Technology And CEO Niu Li said.
Last week, on August 15th, the 4D imaging radar start-up company - Aotu Technology announced the completion of the seed round financing (millions of dollars), in which Li Yifan, the CEO and founder of LiDAR listed company - Hesai Technology, participated as an individual shareholder. The round of financing will be used to accelerate product development and large-scale commercial landing.
One of the characteristics of the company's product solutions is cost reduction.
As we all know, at present, only three suppliers in the Chinese market, ZF, Senstech, and Freetech, have realized the mass production of 4D imaging radar, but the scale of the vehicle is far less than that of the more expensive lidar.
On the one hand, compared with traditional 3D millimeter-wave radars, the technical maturity of 4D imaging radars has not yet reached the "sweet spot" of scale in terms of hardware architecture, software algorithms, and multi-sensing pre-fusion.
To this end, the solution proposed by Aotu Technology is: through the solution of 4 cascaded (12TX+16RX) + low-cost SOC computing chips for cars (replacing the FPGA used by most of the peers), combined with the optimization capabilities of the underlying computing architecture and high-efficiency signal processing algorithms.
progress bar, 6 percent
Video: Benchmarking between Aotu 4D Radar and Hesai AT128 LiDAR
On the other hand, for lidar, car companies tend to give priority to removing lidar as the price of high-end smart driving models drops. "Compared to urban NOA, high-speed scenarios are relatively mature in terms of reliability and safety."
At the beginning of this month, Lixiang’s best-selling model L9 launched the L9 Pro version. Compared with the lidar version of the L9 Max, only the camera and forward-facing 4D millimeter-wave radar are retained, the chip is replaced by the Horizon from Nvidia, and the price of the new car is reduced by 30,000 yuan.
In addition, Weilai Automobile also announced earlier this year that it will cooperate with NXP to jointly develop 4D millimeter-wave radar, so as to enhance the environment mapping and scene understanding capabilities of existing perception systems, and it will not be affected by weather or lighting conditions.
From the data point of view, at present, whether it is L2 with lane change assistance or NOA, how fast the price of the standard model drops, means that the scale effect of intelligent driving can be expected to be released.
According to the monitoring data of Gaogong Intelligent Automobile Research Institute, from January to June 2023, 3.2435 million new passenger cars equipped with L2 (including L2+) equipped with pre-installed standard equipment in the Chinese market (excluding imports and exports) will be delivered, a year-on-year increase of 37.65%; The standard configuration rate is 34.90%.
Among them, non-NOA lane change assistance accounted for 8.34% of the total delivery volume of L2 (including L2+), and the average delivery price of new vehicles was 342,700 yuan; NOA accounted for 6.53%, and the average delivery price of new vehicles was 380,400 yuan;
Previously, Li Yifan said in a public interview that it is more difficult for lidar to enter the low-end electric vehicle market with a price of less than 200,000 yuan, because it is not the main configuration; For electric vehicle products, lidar is the main selling point.
And this time as an individual shareholder involved in 4D millimeter-wave radar, Li Yifan may not only be a market "hedge" for lidar, but another meaning may be to enhance the fusion performance at the 4D perception level.
As the current pacesetter in the mass production and delivery scale of global LiDAR pre-installation, the industrial layout of Hesai Technology and its founders undoubtedly convey the new trends behind the industry.
In fact, in the past year or so, a number of lidar companies have successively launched solid-state lidars for blind spot applications. On the one hand, they achieve larger-scale boarding by reducing costs; Ability to build around perception.
According to Gaogong Intelligent Vehicle Research Institute, the threshold for hardware development and manufacturing of a single sensor has dropped significantly. For example, in the field of millimeter-wave radar, many car companies have also set foot in it. But for the multi-sensor product portfolio, there is still a high threshold.
In the eyes of industry insiders, multi-sensory fusion, if it is a product from different suppliers, there is still a lot of collaborative work in the subsequent system development process. The full-stack perception means that many integrated engineering developments can be front-loaded, which is also a cost reduction for car companies.
For example, for the perception configuration for high-level autonomous driving, in Mobileye's view, the two sets of pure visual perception based on the camera and 4D imaging radar + lidar fusion perception should be paralleled, which is the real redundancy.
First of all, the camera is a passive sensor, regardless of millimeter-wave radar or lidar , it is a dynamic sensor, and it has more advantages in terms of fusion. And more importantly, FMCW .
Taking the current mainstream lidar ranging method (ToF) as an example, FMCW has great advantages in anti-interference, detection distance, solid-state compatibility, millimeter-wave radar compatibility, and whether it contains speed information. FMCW is also the distance measurement and speed measurement method of millimeter wave radar.
This means that since FMCW can directly measure speed information and resist interference, it can realize effective perception fusion of 4D imaging radar and lidar, further improving confidence.
In the prospectus of Hesai Technology, it is written: ToF is the mainstream solution for medium and long-distance laser radar in the current market. In the future, with the maturity of FMCW laser radar and the upstream industry chain, ToF and FMCW laser radar will be on the market. coexist.
At the same time, the application of silicon-based optoelectronics technology to the FMCW system can greatly reduce the size of the optical signal processing module and reduce power consumption, and improve the performance, integration, consistency and reliability of the system. In addition, the mature CMOS process ensures that the integrated solution can be mass-produced at low cost.
In the industry, most people also define FMCW lidar as 4D lidar radar. In addition to traditional lidar that can detect three-dimensional positions, it can also detect the fourth dimension of the instantaneous speed of each scanning point. At the same time, the output raw 4D data provides real-time camera-grade images with up to 1000 lines per frame and no motion blur in static scenes.
For example, Aeva's self-developed silicon photonic chip module integrated laser transmitters, optical devices and receivers on-chip (mature semiconductor process) to achieve cost and volume optimization.
Although generally considered a more complex and less mature technology than current LiDAR solutions on the market, FMCW has gained a growing following.
The same layout of the FMCW lidar track, and Mobileye. As one of the few companies in the industry that deploys 4D imaging radar and FMCW lidar at the same time, it also has its own views.
According to Mobileye founder and CEO Amnon Shashua, the entire lidar industry will move towards FMCW. "At this point in time, this is a difficult but correct choice." In addition to technical performance, cost is regarded as a key factor in the scale of pre-installed mass production.
For example, due to the characteristics of components and system architecture, FMCW lidar is easier to integrate core components into semiconductor packaging (PIC+ASIC) than ToF, and realize wafer-level packaging and testing, thereby greatly reducing the cost of lidar core components (You can refer to the cost curve of the semiconductor industry).
In addition, Gary Hicok, senior vice president of engineering at NVIDIA, said that in addition to instant speed detection, FMCW lidar also has 4D perception, which can provide new functions that traditional lidar cannot achieve, including super-resolution and 4D positioning (can replace the current IMU, reduce the cost of car companies).
4D imaging radar + FMCW lidar is the solution given by Mobileye.
The reason is that, in addition to ranging range, elevation and azimuth sampling capabilities, FMCW also adds direct acquisition of velocity (real-time Doppler velocity measurement like mmWave radar). This means that both can detect and quickly identify small and fast (moving) targets at longer distances at the same time.
And Tesla, which has released its self-developed 4D millimeter-wave radar, its Autopilot hardware design team is also gradually introducing relevant talents in the lidar industry; engineer.
According to Gaogong Intelligent Vehicle Research Institute, as the next-generation sensor technology gradually becomes the mainstream of the market, the diversification of intelligent driving perception combinations is also helping high-end intelligent driving to cover more models in the price range. For sensor manufacturers, it is also necessary to plan in advance.