New Year's Eve! Thank you very much to autolab's readers for taking the time to read our article on the occasion of the Chinese New Year with laughter and laughter, and once again wish you all the best and happy family!
"Year of the Tiger Q&A" is a special plan of AutoLab for the Spring Festival of the Year of the Tiger, we will call the technical experts of lidar, chips, intelligent driving, intelligent cockpit and other leading enterprises in the field of automotive technology, and look forward to the technological development of 2022 together!
In this issue, we greet readers with a senior product technology expert from RoboSense (at the request of the respondents, here is a pseudonym). Leave a message in this tweet and you will have the opportunity to get a New Year's gift from AutoLab and Sagitar Juchuang!
Sagitar Juchuang M1 is the world's first MEMS LiDAR mass-produced and delivered by car specifications, and popular models such as Xiaopeng G9, Zhiji L7, Lucid Air, WM M7, and GAC Aean LX Plus will be equipped with this product. In addition, Extreme Kr, BYD and Lotus have also signed strategic/fixed-point cooperation with Sagitar Juchuang.
The theme of this chat mainly revolves around two questions, what kind of Buff bonus can lidar bring to assisted driving? In what direction will lidar technology develop?
What kind of experience upgrades will lidar bring to intelligent driving?
Lidar is simply able to directly measure the real three-dimensional environment, provide depth information at the same time, but also has a high resolution, on the basis of the camera plus millimeter wave radar, improve the environmental perception of the perception system.
Previously, lidar has been used in L4 self-driving cars such as Waymo and Baidu Apollo, and the high cost and difficulty of passing the car rules are the two main reasons that hinder the landing of lidar in mass production passenger cars.
In recent years, with the maturity of semi-solid-state lidar such as Sagitar Juchuang M1, many passenger cars have begun to carry lidar. When car companies plan a hardware to get on the car, they have corresponding functional definitions, which will directly affect the selection and layout of lidar, and ultimately bring different experiences of assisted driving.
Therefore, even if they are all models equipped with lidar, the experience will be different, and the main thing behind it is related to the definition of the function of the car company. To sum up, at this stage, smart cars are equipped with lidar, and there are three main types of experience upgrades.
The first is to improve the safety of existing ADAS functions. This kind of experience can be improved with more than 1 lidar.
Cameras and millimeter-wave radar cannot ensure that assisted driving is 100% safe, and countless cases have demonstrated this. Lidar can stably output high-resolution data with depth information, which is more advantageous for detecting dangerous obstacles in front of the field of view and predicting moving objects, which also means that loading lidar can improve the safety of existing assisted driving.
In the loading position, we found that mounting on the top will have a longer view, will not be completely blocked by the vehicle in front, and it is easier to play the value of the long detection distance.
Higher resolution, longer detection distance, higher confidence distance and location relationship can help vehicles to avoid leakage recognition of stationary vehicles, pile barrels, etc. on the basis of high-speed pilot auxiliary driving function, and respond in time to avoid accidents.
The second category is the expansion of functions, intelligent driving adapts to more driving scenarios, representative of the city pilot auxiliary function or L3 level traffic congestion automatic driving, for this type of functional car companies generally want 2 or more lidar.
These two types of functions are generally aimed at driving scenes with a speed of less than 60km/h, and need to cover the range of more than 180° in front of them as widely as possible, so the detection distance requirements are smaller, but they are more concerned about the FOV field of view and frame rate. The frame rate can be understood as the timeliness of environmental detection, and the higher the frame rate, the more it can reflect the real road environment in real time.
The third category is the "cost experience" of the L4 autonomous driving industry. At this stage, although robotaxi and roboTruck are still using traditional mechanical lidar, in order to achieve vehicle-level mass production, more autonomous driving companies have begun to plan to replace mechanical lidar with MEMS lidar.
"We feel that since the second half of last year, RoboTaxi's order demand for MEMS lidar M1 has gradually surpassed mechanical lidar. Some of the subsequent models of the Robotaxi enterprise plan may be gradually replaced with a system lidar scheme similar to the M1."
For reasons of commercial efficiency, although mechanical lidar can achieve 360-degree FOV perception, the cost of a single mechanical lidar is much higher than that of MEMS lidar, and manual calibration is required. Therefore, even if 4 to 6 MEMS lidar is used, it still has a cost advantage.
In addition, M1 is a vehicle-grade product, the robustness and reliability of the product is better, which can greatly reduce the operating costs in the later stage and remove a major obstacle for large-scale commercialization.
Therefore, Autolab believes that consumers should be more rational to look at the models equipped with lidar, lidar is not the more the better, the important one is to see the car company's definition of the function of the product, and the other is to see whether such a functional definition is in line with its own automobile consumer demand.
Who is better than MEMS or microtransfers? 1550nm light source is more advantageous than 905nm light source? OPA pure solid state is the future, semi-solid state is the transition solution? Such a dispute over technical routes has also been influencing many people's judgments about lidar.
"The so-called technical route that is argued or tangled up is actually the biggest misunderstanding of lidar itself," Su Xiaotian told AutoLab. "When we come to the industry over the years, we don't think that the technical route itself must be competed for, but more importantly, it is to see how different technical routes meet the needs of customers for pre-installation and mass production."
A few years ago, many people thought that Valeo and Huawei's micro-mirror solutions were unreasonable to do high lines, and for MEMS solutions, there was also a period of time in the industry's cognition that its reliability was questioned. However, sagitar Juchuang, Huawei, Livox and other manufacturers who choose different technical routes have successfully passed the vehicle regulation level certification and have a good performance after getting on the car.
In addition to the scanning method, there is also a laser light source. Since the light source above 1440 nm can be strongly absorbed by the cornea and lens of the human eye, the 1550 nm laser light source can use higher power and exposure time than the 905 nm light source. Therefore, many people will feel that if they want to detect a longer distance, they must use a 1550nm light source.
Su Xiaotian believes that the "light" emitted by the 1550nm light source is very good, but for the light source device itself, Sagitar Juchuang is actually holding a certain skeptical attitude, mainly for the following two reasons:
First, in the process of practice, Sagitar Juchuang found that with the development of upstream components of the 905nm light source, the ranging performance of the scheme still has great potential for development, and it also steadily increases the detection distance under 10% reflectivity to more than 200m.
Second, the temperature rise performance of the 905 nm scheme (performance degradation due to temperature rise) will be better than that of the 1550 nm light source. At the strict on-board high temperature, the actual detection distance of 1550nm is not obvious compared with the 905nm scheme.
Therefore, Autolab found that looking at the technology route cannot separate the application demand and the supply chain environment, and there is no absolute advantage in the relationship route of maturity.
What will be the next product evolution direction of Sagitar Juchuang?
From the interview, we found that sagitar Juchuang's next products will develop together around the horizontal and vertical development of M1. Horizontally, products such as solid-state lidar will be launched to enrich the product line and maximize the needs of the development of intelligent vehicles.
Vertically, the MEMS modular platform based on M1 will be continuously upgraded and optimized to launch a series of products.
Laterally, Sagitar Juchuang will launch a short-distance blindness repair product based on the Flash solution, and there will be prototypes that may potentially meet the B sample in the next few months. The Flash solution has a strong cost advantage, and it is also very convenient for large-scale mass production, but the problem that is difficult to overcome lies in the detection distance, so using Flash to make up for blind radar scheme is the more mainstream practice at present.
AutoLab believes that with the further development of intelligent driving, the requirements for cost control are increasing, and the blind lidar will usher in a large demand explosion, and the installed capacity will be higher than the main radar of the medium and long-range detection distance.
In addition to enriching the product line and gradually improving the key features of the M1 platform, capacity ramping has also been the top priority of Sagitar Juchuang. In the case of supply chain satisfaction, mass production reflects how a product can ensure consistency under large-scale mass production, and the ability to gradually explore the price, which is not less difficult than research and development. When Tesla was in the mass production stage, Musk also issued a similar sigh, and also believed that the manufacturing link of advanced products was more difficult than the design link.