Over the past six months, many investment agencies and car companies have to find a text on laser radar in writing. The problems they mentioned, there are many overlap.
There are many reports on laser radar, but the information fragmentation is very serious, and it is still confused for some basic concepts. Therefore, after answering a large number of homogeneous issues, the author decided to organize these questions, and be a “Popular”.
Q1: Which technology route laser radar is best?
A: This is not a good problem. A more accurate problem should be “which technology route is mainstream at this stage.”
First of all, with regard to the technical route, there is a need to clarify that there is a variety of classification methods in laser radar: according to the distance method; according to the laser transmitter, the detector is divided; according to the scanning method.
According to the ranging mode, the current car laser radar is mainly divided into DTOF (flight time) and ITOF (by measuring phase offset indirect measurement flight time, divided into FMCW and AMCW), the current DTOF is the mainstream market, and FMCW is in Pre-research phase.
The difference between the two is that DTOF is the time difference between the transmitting a bundle laser pulse and the detector to receive the echo signal, directly calculate the distance between the target object and the sensor; and the FMCW is transmitted by adjusting the laser frequency and detects the laser frequency in time. The shooting signal between the return wave, while completing the detection of the distance and speed of the target.
At this stage, TOF is the most mainstream ranging scheme, but companies such as Mobileye, AEVA and Blackmore (acquired by Aurora) have used the FMCW ranging scheme from the beginning. FMCW and TO are almost two independent categories, except for optical lenses, scanning devices are similar, and others are completely different.
Currently, the industry discusss more laser radar technology routes, which are discussed in the TOF ranging scheme.
The structure of the TOF laser radar is divided into three parts: laser transmitter, laser detector, and scanning components. The laser emitter can be divided into 905 nm and 1550 nm according to the wavelength. It is mainly divided into EEL (side emission laser) and VCSEL (vertical cavity emitting laser) according to the integration; the laser detector is integrated and photolithography Sensitivity is mainly divided into APD (avalanche photodiode) and SPAD (single photoelectric detection avalanche), etc .; according to scanning components, can be subjected to mechanical rotation, mixed solid and pure solid state.
(The type listed in each of the above classification methods is incomplete, only the most representative solution is selected.)
Now, the external discussion of laser radar classification, usually specifically, according to scanning components.
Traditional “mechanical rotation”, this concept has no dispute, that is, the product is rotated 360 degrees by the motor; mixing the concept of solid state, simply, there is a part of the component is mechanical movement, the other is not exercise Although the concept is clear, there are many misunderstandings in the past few years; pure solids, there is no mechanical moving parts, which is very clear, but there are some misunderstandings.
During 2017-2019, the mechanical rotary product is given by some old manufacturers, as long as the partial partial part is sealed in a box, it becomes “mixed solid”, but this is actually mechanically rotating. At that stage, MEMS (microscopic mirror) is scheduted as “pure solid” class, which is also misunderstanding because MEMS also has mechanical moving parts.
In 2020, mainstream manufacturers have basically consensced on how to classify laser radar according to scanning methods –
Mechanical rotary: mechanical part (scanning module) and electronics (laser transceiver module) are moving – by the motor with a 360 degree rotation.
Mix solid state: The laser transceiver module is not moving, only the scan module is moving. According to the motion mode of the scan module, the mixing solid is divided into three types of MEMS, transfer mirror and prism (this will be more detailed).
Pure solid: not only the laser transceiver module does not exercise, but also, the scanning module has no mechanical movement. The pure solid state solutions mainly have both OPA and Flash.
Overall, from mechanical rotation to semi-solid, then to solid state, the level of integration is getting higher and higher, and the cost is getting lower.
Mechanical rotary laser radar has a large size, high reliability, difficult to overclot, high price, but its performance index is very strong (long distance, horizontal scan angle). It is mechanically rotating laser radar to help the automatic driving industry completed the stage of 0 to 1 – at this stage, its shortcomings can be inclusive.
Mechanical rotary laser radar is mainly sold to Robotaxi test fleet, due to few demand, Robotaxi customers are not highly sensitive to laser radar prices; and these B-terminal customers have a technical maturity of laser radar, there is a reliability. Comparative expectations, in order to ensure safety, they will also strictly follow the life of the manufacturer, if it is over, will be replaced in time. In addition, they also arrange special inspections and maintenance of laser radar, and have problems can also be found in time.
Currently, the highest laser radar in the Robotaxi market worldwide is the 40th and 64-wire products of Haiyi. According to the information on the Zoai, after 2019, the 64-wire product sold is better than 40 lines. The author once asked several customers, and their prices of mechanical rotary laser radar have high opinions, but their performance is quite approved. However, if you want to put the laser radar on the front load bunter, it is different. Most of these cars are to sell to end consumers to do private cars. Is the consumers who regularly check whether the laser radar on the car has a problem?
Moreover, it is possible that the design life of the laser radar is full, but some consumers are still “not wrong.” In the private car, it is more than 16 hours, and the challenge of laser radar is greater, the challenge of laser radar is greater.
Therefore, we must enter the pre-installation market, and the laser radar must reach the car grade standard.
Starting from 2020 Q4, regardless of laser radar manufacturers or car companies are high-profile “front mass production”, in this stage, in addition to performance indicators, car gauges, integration, mass production, cost, etc. need to focus Problem, this stage, mechanical rotary laser radar basically exits the competition, now, it is basically a mixed solid product.
In the phase of the automatic driving industry from 1 to 10, mixing solid laser radar will play an important role. For example, Huawei does not do the Robotaix test market, so skipping mechanical rotary programs, starting from mixing solid state; in the mechanical rotary laser market, the most competitive Herpet, also mixing the market in the development of the market Solid product (MEMS, transfer).
However, in a longer time dimension, mixing solid state laser radar is only a transitional morphology, and pure solid laser radar is the future. Pure solid states OPA and Flash, but take the lead in proposing QuaNergy from the OPA route, from the automatic driving market, currently, mainstream manufacturers have also developed OPA products; current, IBEO, Mainland, Ouster and other companies pure solid state Laser radar is based on a Flash scheme. Among them, Ouster’s Flash Laser Radar is equipped in a number of unmanned trucks, mines, and sanitation cars.
Even if it is a chassis and Huawei, which is the main mechanical rotary and mixed solid laser radar, the merchants and other manufacturers have also admitted that pure solid represents the future trend, so they all have related technical reserves. He mentioned that the pure solid laser radar of its reserves is not a Flash, but also emphasizes “electronic scan”.
The pure solid laser radar is not only small, but also because there is no moving part, the reliability is higher, and the cost is lower after the technology is mature. However, the fatal weakness of pure solid laser radar is a shorter detection distance, such as IBEO’s Flash Laser Radar IbeoneXT, the detection distance is only 130 meters.
To be a long detection distance, there are two means: improve the power of the laser emitter and improve the sensitivity of the laser detector. At this stage, these two technologies are still immature. These two techniques are mature, and the cost can also be accepted, and Flash is the pure solid state represented will become mainstream.
At the end of 2017, Velodyne executives said in an interview with the author: “The Flash scheme is not suitable for the car, and we have a consensus inside. Now, Velodyne has appeared seriously, and they did not expect, As technology advances, the detection distance of Flash laser radar is capable.
Q2: MEMS, rotation mirror, prism-type representative players which company, where is the specific difference point?
A: The representative of the MEMS program has innoviz, sagitar, pioneer, MEMS laser radar advantage is that the resolution is high, the integration is high, the size is small, the cost is low in existing programs, the lowest cost However, the point of the MEMS product is spit with the spit in the scan mirror, resulting in the average energy of each beam, so that the detection distance is relatively short under the prior art conditions.
However, due to cost and integration, MEMS products have strong competitiveness in the low speed automatic driving market. Sagitar Gongcheng has long binding the rookie this strategic customer, got a large number of real feedback from the customer in more than three years, and the technological progress is very fast.
And, for the L2 + or NOA of the passenger car, because the driving responsibility main body is still a human, the automatic driving system is only a browning, and it is not demanding 100% perfect. Therefore, after the performance and cost, there is also a comprehensive consideration. Car companies choose to carry MEMS laser radar on the front load. Lucid AIR, which is about to carry Sagitar, is a MEMS laser radar, as Lucid’s first quantitative car Lucid Air.
Although the main mechanical rotary laser radar, it was released in the MEMS scheme. According to the information mentioned in the prospectus, the company will also develop MEMS products.
The representative players of the rotation mirror include Fareo, Luminar, Innovusion (Laser radar supplier of 蔚 来 ET7) and Huawei. The first laser radar, the laser radar, is Based on the rotation plan, because this program is easy to pass the car, and the manufacturers of other main pre-installations have also followed. (There are many reports and research institutions to classify Luminar and Huawei’s technical lines as MEMS. This is a misunderstanding. Every company will reserve a lot of technologies, and they apply for relevant patents for some technical programs, but actually No product. Some research institutions only judge the technical route of others products through patents, but they have not directly contacted relevant companies, so deviations appear on the analysis results.)
The difference between the transfer and MEMS is that the scanning mirror of MEMS is vibrating around a diameter, while the mirror is rotating around the center. This scanning method means that the power consumption is relatively low, the heat dissipation is difficult, and it is easy to achieve a relatively high reliability.
At present, in the header of the mass production order has been taken, it is relatively relatively relatively relatively large. Although the mechanical rotary products are very competitive in the Robotaxi market, in fact, the company has been layout in 2017, it is only relatively low-key, and has not been disclosed.
In addition, the prospectus of Herbi shows that the product Pandar ST in front of the premium market is also used.
Two-dimensional mirror (IBEO, Hi-race) and two-dimensional mirrors, so-called one-dimensional mirror, only one side scan mirror, two-dimensional mirror, one vertical and horizontal scan mirror. How many laser transmitters are there in a single-dimensional mirror, which means not only cost high when doing high-line products, but also is difficult to integrate, so the number of lines is difficult to do (Fareo Scala 1 Only 4 lines, Scala 2 is only 16 lines, the reason is this); the two-dimensional mirror is similar to MEMS, which is only a small laser transmitter, reflecting refraction and reflection in the high speed of scanning mirror. ” Multi-line “effect, not only can save the cost of lasers, but also high” line counts “.
The Pandarst of Haiyi will use a one-dimensional mirror scheme, but it is different from Scala, and the Hero Plan integrates the laser transmitter and detector into the chip (this goal will be implemented in the 2.0-stage phase of the Wope chip size ROADMAP) to achieve laser One of the transceivers and “line counts”, which can not only achieve a high number of lines based on one-dimensional mirror, but also a dividend of molar law.
The representative of the prism-type scanning scheme is Livox. The prism type scheme is used by non-repeating scanning techniques, which means that the algorithm that needs to be matched is different from the mainstream of laser radar based on repeated scanning technology.
Compared with MEMS and two-dimensional mirror scheme, the number of laser emitters for prism is more, and also achieves a higher point cloud density, further detection distance, but due to the motor speed ratio rotation mirror scheme High and double, prism schemes have a very big challenge of reliability such as motor bearings. Livox has accumulated precision motor manufacturing technology while doing drones, and confidently overcomes this difficult, and other manufacturers have freshly explore this technology.
Livox’s prism products can also enjoy a “dividend”: Livox’s brothers “Dabin Car” will ask them to feedback, so they have formed a closed loop, which is conducive to the rapid progress of technology.
Q3: Which of the MEMS, transfer mirror, and prism types is better in mixing solid state?
A: At present, the technical route of laser radar is still in the stage of hundreds of people. These three programs have mainstream manufacturers attempting, even, the same company is doing two or more programs at the same time; and there is no product On a large scale, there is no verification. Now the evaluation “Who is not working”, it is premature.
For example, even if you got a $ 1.5 billion orders as early as one and a half ago, according to the information published before the listing, only 100 laser radars were sold in 2020, and a large amount of delivery would wait until 2022.
Moreover, many orders published by LumianR may be just a supply agreement, mainly to restrain the price of suppliers, and there is not a strong binding force on the order size of the vehicle enterprises. The following is the list of the clients in the PPT of the investor before listing, and the Audi, General, Ford, and the awareness, but in fact, Audi is mainly working with Ibeo, universal Cruise and Ford ARGO They are all self-developed laser radar (Cruise is also working with the game), and it will choose INNOVISION.
The original words of a laser radar manufacturer are: “Even if it is signed with a car company, there may be variables in subsequent cooperation, and may not cooperate for a long time; now there is no contract, no chance “
The author estimates that which technology route and which company’s product is more powerful, there will be a relatively clear answer before 2023.
Q4: Does this industry have a first advantage? For example, if the MEMS scheme is the best, the other routes are wrong. If you want to change, are you going? A: The first choice needs to clarify, Haiyi and Huawei have MEMS technical reserves, even if there is no related products, there are patents, so they don’t have the “I don’t know” after the MEMS route wins. “
Moreover, MEMS is also transferred or prism, and the laser transceiver module can be universally, which is mainly in the scanning scheme. Then, a company can quickly switch from a transfer or prism scheme to a MEMS program, depending on the question: the true barrier of the laser radar in the scanning module or in the laser transceiver module?
If the former, it is difficult to switch to another scanning scheme. If it is the latter, it is not difficult.
Some vendors believe that the scanning plan is a point of different competition, but the Hero mentioned that the author took the author in July, the scanning program was only the “external power” of the laser radar, and the laser transceiver technology is “internal strength”. Herbai believes that the laser transmission and receiving technology accumulated in Mechanical rotary products can be reused to transfer, MEMS and pure solid products, only need to fine-tune individual parameters. OUSTER also holds similar views.
(Laser radar core technical logic block diagram; from Haiyai Book)
Last year, when the cost of writing in the Luminar Prosebut, mainly emphasized how much the hardware cost of the ASIC chip, a laser transmitter, and a laser detector, but did not mention the cost of the mirror.
The author’s understanding is that after Fareo has proved that the transfer scheme is more likely to pass the car, many manufacturers are aimed at the transfer mirror when developing the products of the premium market, which means the barriers of scanning schemes. It may be not so high; in fact, doing long-probing distance, increasing integration, often focusing on the laser transceiver module “Being article.”
In the pure solid stage, the laser transmitter and detector are integrated into the chip, and no separate scanning parts are required. At this time, the laser radar is highly dependent on semiconductor technology, and it is difficult to scan the scanning scheme to account for much weight.
Therefore, in terms of scanning schemes, the first advantage may not be obvious; but in terms of laser transmission and reception technology, the first advantage may be more obvious.
Q5: Which piece is the highest in the cost structure of laser radar?
A: From the core component components of the current laser radar, the chip is the key to the grade, including control chips, ASIC chips, lasers, beam control mechanisms, and photodetectors. Luminar is shown in the present invention, which has been locked with the supplier’s protocol, and the three costs can be controlled within $ 100 under the premise of the order amount reaches a certain scale.
Q6: Is the laser radar chip need self-research?
Answer: Early, there are many manufacturers used by the manufacturer, such as the French market research institution Yole’s report shows that Gao Tong, Texas Innotek and 理光 are providing chips for laser radar manufacturers. However, due to the laser radar is a relatively new product, the technical route is still uncertain. Suppliers have no mature solutions, so the difficulty of external cultivation is not low.
Moreover, the early barriers in the laser radar in optical knowledge, but from the long run, how to improve performance, how the cost is reduced, is a semiconductor problem. The chip is the lifeblood of laser radar manufacturers. The laser radar manufacturer does not have the self-research ability of the core chip, it is difficult to survive in this market. Therefore, Luminar determines the self-developed laser radar chip from the beginning, and OUSTER, which will be listed in the United States has also been self-developed ASIC chips on the Flash laser radar.
Q7: With the evolution of technology, is there a big change in the function of laser radar?
A: The basic function will not change, but the performance will be better, mainly reflected in: the detection distance will be more and more, the ranging accuracy will get more and more higher, the resolution will get more and higher.
Q8: There are some laser radar manufacturers to provide a sense of knowledge, doing data pretreatment, so that the laser radar output is no longer the original data, it is said that this can sell a higher price, then, the car company is more Like this scenario with a sense algorithm, or only buy a laser radar hardware?
A: This is a very good problem. Luminar and Sagitar Gongca have clearly put the algorithm as an income source. It also mentioned in the Zo-raising book that there is almost 150 million yuan in the financing amount to be used for algorithm development. But it is more like to prefer the solution, and different car companies will have different options.
Xiaopeng, Wei, Ideally, this kind of car new force and the Great Wall have the traditional car enterprise that must master the core technology. It is basically unlikely to use the algorithm provided by laser radar manufacturers; but many half will not I have a sense of knowledge, and I am rushing to use the traditional car enterprise of laser radar, I will have to rely on the perceived algorithm provided by the laser radar manufacturer. After all, consumers only bring the experience of laser radar, whether the money is spent , And not care about your laser radar algorithm is your own or supplier.
In the next few years, there are still many laser radars selling to the car road collaborative project, these customers are engaged in infrastructure companies, they must have failed to do algorithms. Laser radar manufacturers have to do business with these customers, not what you want to provide algorithms, but you have to do this, if you don’t provide algorithms, this market you can’t get it. Of course, the algorithm of laser radar is much more simpler than the camera algorithm, so manufacturers provide algorithms to increase the added value of the product, but the algorithm itself is not sufficient to constitute a competitive barrier.
Q9: Huawei’s laser radar is bundled with MDC?
A: Will not bundle. The MDC team has frequently contacted with other laser radar manufacturers. I hope that MDC interface is compatible with laser radar of other vendors. This is a better practice for Huawei, other laser radar manufacturers and car companies are a better practice.
Q10: Where can I provide a full stack of programs in this way? Where is the survival space of Sagitar?
A: The capacity system and demand of each household factory are not quite the same. Some main units will tend to adopt Huawei’s full stacking program, but there are also many host plants to tend to self-research software algorithms, according to algorithm capabilities and demand to select chips. Sensor supplier.
Q11: I am from the strategic investment department of the car enterprise, and if the car company is necessary to be self-developed laser radar?
A: From a long time, the difference in automatic driving ability between various cars is in the decision-making algorithm, not the perceived link – “Just like we evaluate a cow is not forced, it is to see his thoughts. , Not to see his vision cow, “(Guo Jiyi), and after the technology matures, the perception is a general technique that is relatively high.
From the perspective of technology, invest in a general technology, helplessness to improve the competitiveness of the vehicle enterprises. However, if the purchase volume of the car enterprise is large enough, it is another thing to enjoy the discount price through investment.
Q12: As a car company, do we have a laser radar company, accounting for 10-15% of our shares and a board seat, and strive for the right to give priority supply?
A: Maybe, in a few years later, “priority is equipped” is not that important.
In the previous year, in the stage of Velodyen monopolized, the production capacity was extremely short-term, Baidu Ford invests in the purchase of Veldoyen’s laser radar, but now, the laser radar is already a full competitive market, with a total of more than 100 players There is a influentiality of about 10, after a few years, production capacity should also get up, “priority” must not exist?
Moreover, if your algorithm has no cattle, it is necessary to match a very advanced sensor, the meaning of “priority” is not big?