As a dozen leading companies are trying to create a powerful and inexpensive lidar

Lidar is essential for robomobiles - and this is how some of the leading sensors work

Lidar , or light radar, is a technology critical to creating robomobiles. Sensors provide the computer with a three-dimensional point cloud that indicates the space surrounding the car, and its concept helped teams win the 2007 DARPA Urban Challenge . Since then, lidar systems have become the standard for robomobiles.

In recent years, dozens of startups have been created that work with lidars and compete with industry leader Velodyne. All of them promised more reasonable prices and improved work efficiency. In 2018, Ars magazine already made a selectionthe main trends in the lidar industry, and described why experts expected improved and less expensive systems in the next few years. There were no details about the companies themselves in that article, mainly because they kept information about their technology secret.

But over the past year I have received a continuous stream of advertising coming from the developers of lidars, and chatted with a large number of their representatives. Ars is in contact with the directors of at least eight of these companies, as well as industry analysis companies or their clients. All this communication allowed us to form a good idea not only about the trends in the lidar industry, but also about the technologies and business strategies of individual companies.

Today, there are three main differences between lidars. After describing these features, it will be easier to understand the technology of nine leading lidar companies.

In order not to inflate the article in vain, we describe independent companies that are mainly engaged in lidars. Therefore, we will not describe our own lidar technology from Waymo, startups working with lidars that GM and Ford bought for themselves in 2017, or attempts to develop lidars from larger companies such as Valeo (which made lidar for Audi models 2018 and 2019 A7 and A8), Pioneer or Continental. It is difficult to get the details of their technologies from these large companies, but even without them there is something to describe.

Three major factors that distinguish lidars from each other

The basic idea of ​​the lidar is simple: the sensor emits laser beams in different directions, and waits for their reflections to return. The speed of light is known, and the travel time back and forth gives an accurate estimate of the distance.

Although the basic idea is simple, the details complicate things very quickly. Each lidar manufacturer must make three basic decisions: how to direct the laser in different directions, how to measure the time to the round trip, and what frequency light to use. We will consider each of them in turn.

Beam control technology

Most leading lidars use one of four methods of directing laser beams in different directions (two companies, Baraja and Cepton, reported using other technologies that they did not explain):

• Rotating lidar. Velodyne created the modern lidar industry in 2007 by introducing a lidar that housed 64 lasers vertically, and this whole thing rotated at a speed of several revolutions per second. High-end sensors from Velodyne still use this technology, and at least one of its competitors, Ouster, has done the same. The advantages of this approach are 360-degree coverage, but critics raise questions about whether it is possible to make a cheap and reliable rotating lidar suitable for the mass market.
• A mechanical scanning lidar uses a mirror, redirecting a single laser beam in different directions. Some companies use an approach called the Microelectromechanical System (MEMS) to control the mirror.
• Active phased array antenna uses a number of emitters that can change the direction of the laser beam, adjusting the relative phase of the signal between adjacent transmitters. We will describe this technology in detail in the Quanergy section.
• Flash-based lidar highlights the entire area at once. Existing technologies use one wide-angle laser. The technology has difficulties with large distances, since only a small part of the laser light reaches any point. At least one company, Ouster, plans to create a multi-laser flash, in which there will be an array of thousands or millions of lasers aimed in different directions.

Distance measurement

Lidar measures the time that light takes to reach an object and reflect on it. There are three easy ways to do this:

• Travel time. The lidar sends a short impulse and measures how much time will pass before fixing the returning impulse.
• Frequency Modulated Continuous Lidar (LRCH). It sends a continuous beam of light whose frequency is constantly changing over time. The beam is divided into two, and one of them goes to the outside world, and then, upon return, combines with the other. Since the frequency at the beam source varies continuously, the difference in the path of the two rays is expressed in terms of the difference in their frequencies. The result is a picture of interference, the beat frequency of which is a function of time in transit (and, consequently, of distance). This path may seem unnecessarily complicated, but it has a couple of advantages. The NIDM lidar is resistant to interference from other lidars or from the Sun. LIDAR NICHM can also use Doppler shift to measure the speed of objects, and not just the distance to them.
• Amplitude Modulated Continuous Lidar (NIAM) can be seen as a compromise between the two previous options. Such a lidar, like a simple sensor that measures travel time, sends a signal, and then measures the time that it took him to reflect and return. But if simple systems send one pulse, the NIAM lidar sends a complex scheme (a pseudo-random stream of digital zeros and ones). Proponents of the approach say that due to this, the NIAM lidar is more resistant to interference.

Laser wavelength

The lidars described in this article use one of three wavelengths: 850, 905, or 1550 nm.

This choice matters for two reasons. One of them is eye safety. The fluid inside the eye is transparent to light with a wavelength of 850 and 905 nm, which allows the light to reach the retina. If the laser is too powerful, it can cause irreparable harm to the eye.

On the other hand, the eye is opaque to radiation with a wavelength of 1550 nm, which allows such lidars to work at higher power without harming the retina. Increased power allows you to increase the range.

So why do not everyone use lasers with a wavelength of 1550 nm in lidars? Detectors operating at frequencies of 850 and 905 nm can be created on the basis of inexpensive and widespread silicon technologies. To create a lidar with a wavelength of 1550 nm, exotic and expensive materials, such as gallium indium arsenide, are required.

Although 1550 nm lasers can operate with greater power without causing a threat to the eyes, such power levels can lead to other problems. At CES in Las Vegas this year, one man announced that a powerful 1550 nm laser in AEye’s lidar had ruined his camera . And, of course, lasers of higher power consume more energy, which reduces the range and energy efficiency of the machine.

Given all this, let's look at the top ten leading lidar developers.

Velodyne

Three Velodyne products: Alpha Puck, Velarray and Veladome Beam

control : rotation.

Distance measurement : travel time.

Wavelength : 905 nm

Velodyne invented the modern three-dimensional lidar more than ten years ago, and has since dominated this market. The company's characteristic rotating lidars are often used in robotic vehicles, and the company is likely to remain the market leader in 2019. However, some observers are wondering if the company will be able to maintain its leading position in subsequent years.

At the end of 2017, the 64-laser Velodyne flagship lidars sold for $ 75,000 apiece. Velodyne introduced a new model with 128 lasers, which is rumored to be even more expensive - $ 100,000.

Regarding these figures, the representative of Velodyne replied: “We do not disclose the cost of production, however, the announced prices are typical for single products. In purchases of automobile scales, prices are significantly lower, and we are actively supplying car manufacturers with products at low prices. ”

Velodyne also sells less expensive lidars, including a 16-laser washer , which last year sold for $ 4,000. Velodyne also sells a solid model, Velarray. Velodyne says it is a system with a wavelength of 905 mm "with a proprietary beamless friction control method." Velodyne expects that in bulk this model will end up costing less than $ 1,000. However, these lidars do not give such a high-precision result as rotating models with 64 and 128 lasers.

Some critics claim that Velodyne had difficulty in manufacturing and product quality.

“The delicate moving lidar sensors, which are the company's livelihoods, have proven difficult to produce efficiently and with high quality, and they can be annoyingly fragile when used in automobiles, ” journalist Ed Niedermeier recently wrote , citing sources in the robomobile sector.

A company representative argued with such a recall, arguing that Velodyne "over the years has brought the science of manufacturing these sensors in large quantities to perfection", and that "it has been proven that they withstand harsh operating conditions in cars."

Velodyne recently signedlicensing agreement with Veoneer, a well-known company in the automotive parts supply chain. Veoneer has extensive experience creating components that meet the quality standards of automotive companies, and she may have ideas on how to make changes to the classic Velodyne design in order to improve the quality and reduce the price of the product. However, they need to act quickly, as a number of other companies have already set their sights on the leader.

Luminar

Beam control : mechanical scanning

Distance measurement : travel time.

Wavelength : 1550 nm

Many consider Luminar to be one of Velodyne’s main rivals. The company has been engaged in this business since 2012, and last year began the production of lidars in large quantities. The company claims that the quality of its products is at the highest level.

In particular, this is due to the fact that Luminar decided to use lasers with a wavelength of 1550 nm. Using an eye-safe wavelength allows Luminar to twist the laser power, so the lidar sees further. But 1550 nm lasers mean that Luminar has to use exotic gallium-indium arsenide to detect returning pulses. It should be expensive, but Luminar last year informed us that the cost of the receivers in their lid is only $ 3.

Last year, in response to our inquiries about Luminar, Marta Hall, president of Velodyne, pointed out to us a serious drawback of Luminar lidars - high energy consumption. This is especially important because Luminar lidars are fixed sensors with a 120-degree field of view. This means that to ensure viewing of all 360 degrees, you will need four devices from Luminar (taking into account the imposition of their fields of view), instead of just one from Velodyne or Ouster. However, then in a letter, a Luminar spokesman replied that the latest version of their lidar significantly reduced energy consumption compared to earlier models, and consumes "about a circle of about 50 watts."

Luminar also does not say anything about prices. Last May, Luminar Director Austin Russell told us that their lidar would have to “drop in price to a few thousand dollars” in order to compete in the consumer market, and that this issue is “not a problem” for the company. However, from this it follows that at that time the cost of devices was much higher than several thousand.

Luminar is ahead of many lidar manufacturers in real-world deliveries since it began mass production more than nine months ago. Over the past 18 months, Luminar has managed to enter into partnerships with Toyota , Volkswagen, and Volvo .

In a recent interview, Russell pointed to these deals, calling them the company's biggest competitive advantage. He told me that the largest companies are developing luminar-based lobars from Luminar, and it will cost them a lot to switch to competitor products in the future.

Aye

Beam control : mechanical scanning

Distance measurement : travel time.

Wavelength : 1550 nm

AEye has a lot in common with Luminar. It uses a mechanical scanning mirror to control the beams. It uses an eye-safe laser with a wavelength of 1550 nm, allowing it to operate at high energy levels. As a result, the AEye lidar has impressive range characteristics. AEye says that their lidar can see at distances up to 1000 m - this is much more than the 200-300 m that the most expensive devices boast of.

In a December interview, AEye Director Lewis Dussan touted the high-energy impulses that AEye lidar fiber lasers can emit. He said many competitor lidars are based on diode lasers, "limited to 100-150 watts. Fiber lasers can reach up to 100,000 watts - a very short pulse, a large amount of signal. "

Great energy allows you to increase the distance, but it also has its drawbacks. This year, at CES in Las Vegas, one person told Ars magazine that his expensive camera was ruined when he took a photo of the lidar from AEye. The eyes are filled with a liquid impermeable to waves with a length of 1550 nm. But the cameras are not. Apparently, a powerful AEye laser hit the fragile matrix of the camera.

In a statement to Ars, AEye described camera damage as an industry-wide problem. But Angus Pakala, director of rival Ouster, argues with that. He wrote: “Our sensors are safe for the eyes and cameras. And the point. " Luminar said that “we conducted comprehensive tests with the same camera with the same lenses and the same settings as the damaged CES, and were unable to harm it” using the luminar from Luminar.

Most lidars use a fixed scanning scheme. Lidar AEye takes a different approach, which the company calls “mobile scanning”. AEye scan scheme can be programmed and changed dynamically. According to Dussanne, the movable scanning circuitry works with the flexibility of a fiber laser. “From shot to shot, you can control the energy of the pulses,” he told Ars. The software manages not only when the next measurement will take place, but also how much energy will be used - and, therefore, how much distance will be measured next time.

As a result, when the lidar notices a far-reaching object, it can increase the scanning resolution and energy level in this part of the image, and get more data points. The result may be a high-resolution scan that will help to distinguish between a pedestrian, a motorcycle or bulky debris left on the road.

On the other hand, there is a danger of over-optimization. If the lidar spends a lot of time scanning already recognized objects, there is a danger that too little time will be left for systematic scanning, because of which it will skip other objects.

Uster

Beam control : rotation

Distance measurement : travel time.

Wavelength : 850 nm

At first glance, the Ouster lidar looks very similar to Velodyne. These are rotating systems that measure the time of pulses in transit, and both companies sell devices with 16, 64 and 128 lasers. And this is not a coincidence: Ouster specifically designed the product so that it could be used to replace Velodyne instruments, as many potential customers got used to their classic form factor.

But if you open the devices from Ouster, it turns out that inside they look completely different. The classic Velodyne design, judging by the patent, uses 64 separate lasers and 64 separate detectors. Ouster also figured out how to pack 64 lasers on one chip, and their second chip contains 64 sensors that recognize reflected light. Such an integrated design can dramatically reduce the cost and complexity of lidar production.

The most difficult Ouster lidar, due to ship this year, is the OS-2, a 64-laser unit for $ 24,000. Ouster says its range is comparable to Velodyne’s most expensive lidars. Ouster also sells lidars and with a smaller range for just $ 3,500.

Ouster can push 64 lasers per chip usingsurface-emitting laser with a vertical resonator (VCSEL) - in contrast to conventional laser diodes emitting in a plane parallel to the surface. Since VCSEL emit perpendicular to the surface of the substrate, many lasers can be placed on a semiconductor chip. The technology has long been used in user applications such as computer mice, but it has always been considered not powerful enough for use in lidar. Ouster says they figured out how to create a high performance lidar with VCSEL.

Ouster uses another semiconductor technology, single-photon stage diodes(SPAD) to detect returned light. Like VCSEL, SPAD can be manufactured using standard silicon chip manufacturing techniques, and many SPADs can be shoved into a single chip. Thanks to this, it was pretty easy for Ouster to switch from 64-laser devices last year to 128-laser ones, the announcement of which took place in January, and deliveries will begin in the summer. The company just had to replace the chips with 64 lasers and 64 detectors in the old model with the new 128th chips.

And upgrading from 64 to 128 lasers is just the beginning, says director Angus Pakal. He expects that in a few years the company will introduce lidars, which will have at their disposal thousands - and possibly millions - of VCSEL lasers and SPAD detectors.

So far, Ouster is focusing on creating one-dimensional arrays of lasers for use in a rotating sensor, similar to devices from Velodyne. But Pakala says that the same technique can also be used to create two-dimensional arrays of lasers and detectors - like the matrix in a camera. This can lead to the creation of a new class of flare-based lidars, where each “pixel” will serve its own laser detector pair. As a result, the lidar will have the advantages of a flash - no moving parts, the ability to perceive the "frame" immediately and in its entirety - without sacrificing the range of an ordinary lidar.

The essence of the Ouster strategy is to take advantage of the industrial base of consumer electronics, in which VCSELs are already used in computer mice, for rangefinders on smartphone cameras, and in other areas. Pacala claims that there is still room for improvement in VCSEL in terms of parameters such as brightness, cost and energy efficiency. And all the improvements in VCSEL (and SPAD) technology will automatically work in the hands of Ouster.

Blackmore

Beam control : mechanical scanning.

Measurement of distances : continuous radiation with frequency modulation.

Wavelength : 1550 nm

Like Ouster, Blackmore hopes to take advantage of the extensive infrastructure of the semiconductor industry. However, she is interested in the optical communications industry, not consumer electronics.

At first glance, lidars and optical communication devices are different from each other, but in reality they have more in common than one could imagine. They send information encoded in the light, capture the light later and extract information from it.

“The Blackmore optical layer is based on standard components for fiber optic communications,” the website says .company. “Using decades of fiber optic solutions, we are confident that our designs are scalable and reliable.”

In almost all other aspects, Blackmore lidar is surprisingly very different from Ouster and Velodyne products. Instead of rotating 360 degrees, the lidar is fixed with a field of view of 120 degrees horizontally and 30 degrees vertically. It uses cw radiation with frequency modulation to measure distances, which makes it possible to measure the speed of objects.

Blackmore introduced a new interesting lidar at CES a few weeks ago . Its initial cost is $ 20,000, and it has impressive features. The company hopes to gradually reduce the cost of lidar over time.

Baraja

Beam control : spectroscopic scanning.

Distance measurement : continuous emission with amplitude modulation.

Wavelength : 1550 nm

Baraja - one of the most unusual startups that I talked about last year - and one of the most mysterious.

Most lidars have a field of view of 120 degrees or less, which means they need to buy at least four to ensure full coverage of 360 degrees. This can be expensive, and also requires the placement of fragile electronics on the edges of the machine, where it is very easy to damage.

Baraja's idea is to move all the fragile electronics into the trunk. The signal processor located there is connected via fiber optic to four cheap and durable sensor heads that can be placed outside the machine.

In an interview last summer, director of the company Federico Collarte told me that the four sensor heads “essentially consist of silicon glass. They are cheap, reliable, and withstand the elements well. In the event of an accident, you just need to replace the sensor head. ”

Attractive idea. The problem is that I cannot figure out how it will work - and I could not convince Collarthe to explain it to me in detail.

Baraja describes his lidar as a “spectroscopic scanning lidar,” which means that the laser beams are controlled by changing the frequency of the light passing through the prism. It’s easy to imagine how one can control such a beam in one dimension, but it’s difficult to understand how to achieve two-dimensional control. ”

When I asked Collart about this, he said:“ For the second dimension, we use the same concept of spectral scanning. And we still have an auxiliary mechanical system. "

He added that this system does not include either mirrors or rotating lasers. He said that it" uses the same prismatic optics - we still keep this moment a secret. "

Also, Baraja remains the only company we talked to using amplitude-modulated continuous radiation to measure distances. Collarte told us that one of the advantages of this approach is that "for individual pulses, high energies are not required." Some optical components can be damaged due to power surges, and their absence gives engineers the flexibility to use a wider range of options - which could potentially create a less expensive and more reliable technology.

Collarte says Baraja (like Blackmore) is trying to “transfer components and technology from optical telecommunications”, where large economies of scale help keep product costs low. Baraja seems to be in the early stages of commercialization, but Collarte says the company expects to reduce its cost to “a few hundred” dollars in the production of hundreds of thousands of devices.

Quanergy

Beam control : Active phased array antenna.

Distance measurement : travel time.

Wavelength : 905 nm

Three years ago, Quanergy began to make a big fuss when it announced the creation of a solid product with a cost of less than $ 250, which can be achieved with large-scale production. But critics say the company failed to deliver on its promise.

“Quanergy seems to have a hard time getting the sensors to work at the right distances,” said Sam Abulsamid, an analyst at Navigant, in an interview.



Quanergy is one of the few companies making lidars using active phased array technology. As indicated in the explanationto the 2017 concept:

Phased array is a series of transmitters that can change the direction of the electromagnetic beam, adjusting the relative phase of the signal from one transmitter to another.

If all the transmitters synchronously emit electromagnetic waves, the beam will go straight, i.e., perpendicular to the array. To deflect the beam to the left, the transmitters shift the phase of the signal sent by each antenna, and the signal from the transmitters on the left is behind the signal from the transmitters on the right. To deflect the beam to the right, the lattice performs the opposite action, shifting the phase of the leftmost elements forward relative to the right.

This technology has been used for decades in radars, where radar antennas serve as transmitters. Optical phased arrays apply the same principle to light, packing an array of lasers on a fairly small chip.

If Quanergy managed to get this technology to work well, it would have a ton of advantages. With no moving parts, a solid-state device could be cheap, reliable, and versatile. Quanergy’s lidar, like AEye’s instrument, is programmable and dynamically switches between resolution and refresh rate.

But Quanergy does not have much success in the market. In a November interview, director Luay Eldad said that "we are going through the right steps, we are on schedule." But there is reason to doubt it. For example, Angus Pakala was a co-founder of Quanergy before leaving and founding Ouster in 2015.

Abulsamide points to Quanergy's recent interest in the use of lidars in industrial safety - in this application area, distances such as robomobiles are not required. Eldada told me that Quanergy now has a more typical mechanical-guided lidar designed for the security market.

Cepton

Beam control : proprietary micromotion technology.

Distance measurement : travel time.

Wavelength : 905 nm

Fully automatic robomobiles are the most demanding area of ​​application for lidars, and so far I have mainly described products aimed at this market. But Cepton is an example of a respected lidar manufacturer, mainly aimed at using their technology in advanced driver assistance systems (ADAS). Today's ADAS systems use radar and cameras for lane control and dynamic cruise control. But everyone expects automakers to see lidars on cars of the future that can provide more sophisticated ADAS systems.

The problem is that, as we saw, the best lidars cost tens of thousands of dollars, and this situation may not change even when they are produced on an industrial scale. Therefore, companies such as Cepton are aiming for the production of medium-range lidars that are sufficiently accessible to be included in cars that will be produced in a few years.

And when I asked Cepton Director June Pei about the long-range lid required for robotic vehicles, he disowned the market, saying that he did not think that customers would begin to request such devices in large quantities "in the foreseeable future."

Instead, Cepton focused on the ADAS market, where it is already starting to make deals on large volumes of supplies. Cepton claims that its competitive advantage is price.

“We are the only company capable of selling lidars for less than $ 1,000,” said Pei. Last summer, Cepton announced a deal with Koito, a Japanese company and one of the world's largest suppliers of car headlights, according to which it will include their lidar technology in headlight design. This means that if the automaker decides that the Cepton lidar suits them in all respects, he will be able to easily add this opportunity to his cars.

Pei told me that the micromotion technology that controls the beam is unique to this industry. Traditional MEMS use a tiny mechanically moving mirror to redirect light. But Pei says Cepton uses "a very proprietary optical design that eliminates the mirror, but is still capable of producing a high-resolution image." He also described it as “a small vibratory system that works on the principle of a speaker” - but refused to reveal the details.

Innoviz

Beam control : mechanical scanning.

Distance measurement : travel time.

Wavelength : 905 nm

Innoviz, like Cepton, mainly focuses on high-volume transactions with automakers. It sells affordable mid-range lidars suitable for use in ADAS. And very successful.

Last April, BMW announced plans to install Innoviz lidar in its cars in the 2021 model year. Also involved in this partnership is Magna, a well-known supplier who will help with the logistics needed to install finished parts in thousands of cars.

Automakers are experimenting with many lidar technologies, so many of their manufacturers can boast of making deals with OEMs. But the BMW deal sets Innoviz apart from other competitors - BMW, apparently, is serious about installing their lidars in cars for sale, and not just buying these devices for prototype testing.

In the production of cars, the timing for the development of new products is very long, so Innoviz will have something to do in the next few years, and, of course, one deal will allow Innoviz to conclude new deals in the future. He is optimistic about this deal. "

The deal with BMW, apparently, will be used to implement ADAS, but Innoviz has ambitions in the field of robot cars. In its latest model, InnovizOne The company boasts a range of up to 200 meters with objects with 50% reflectivity and a 120-degree field of view.