Lei Feng Net: On November 3, 2007, six cars made history. They succeeded in completing the autopilot test at a test site simulating a city environment without violating California traffic regulations. It is worth mentioning that, of the 6 cars, 5 are equipped with a new lidar sensor, and to provide the device is Velodyne, but it was an audio equipment manufacturer.
Ten years later, Velodyne continues the legend, whose products are still at the heart of autonomous vehicles. One thing that has not changed, however, is the price of LIDAR, which has only gone from very expensive to quite expensive and far removed from the goal of civilians.
Therefore, Velodyne and a large number of emerging laser radar start-ups now have the main goal of changing this awkward situation.
In this article, we will conduct an in-depth analysis of Lidar technology. In addition to the working principle of laser radar, this article will explain the challenges it faces, after all, to meet the commercial requirements of autonomous vehicles is not easy.
Some experts believe that the key to building a low-cost Lidar is converting a traditional rotary mechanical design to a solid-state design, which allows a significant reduction in the number of moveable parts. Lidar is not only simple to structure and mass-produce but also low in cost It's Having said that, no one knows how long it will take to build a low-cost, vehicle-grade lidar.
The good news is that all experts are positive about this newspaper. To this end, they also cited a large number of examples of technological advances, such as calculators and anti-lock system, their prices are high enough to be very cheap, by mass production. Obviously, Lidar will repeat the same path, that is, the price of Lidar will not be a stumbling block to the popularity of autonomous vehicles.
Laser Radar Urgently Pioneer
* Velodyne's 64-line Lidar Brings Dawn to Automated Driving </ em>
60s of last century, scientists began to use laser rangefinder. At that time, a team at the Massachusetts Institute of Technology measured the distance between the earth and the moon with a laser. However, when it comes to car lidar technology, David Hall, inventor of Velodyne, has to be mentioned.
Hall at the beginning of the 21st century was the CEO (and founder) of Velodyne, a company that focused on audio equipment. But privately, Hall is a real robot enthusiast.
"Our team then boarded the arena for BattleBots and Robot Wars, both robotics, on the pretext of advertising Velodyne's audio," said Hall in an interview with Connection magazine.
As a result, Hall decided to participate when DARPA, the Internet-developed division of the Defense Advanced Research Projects Agency, rebuffed a self-driving car race called the Grand Challenge.
In 2004, Hall and his own brothers were "on the battlefield," and the two trucks they installed were fitted with cameras. However, such a crude configuration can not help them win, or even complete the game. Unfortunately, when the game did not have a car to the last laugh.
In the second year after the game, Hall two brothers abandoned the camera program, in turn, used the lidar. Other teams also switched to LIDAR, but the products on the market were still quite primitive.
At that time, the most popular LIDAR was the SICK LMS-291, a 2D LIDAR system, meaning that it had less than 2.0 "eyesight" and could only "see" a portion of the world.
The system can help the car detect similar obstacles such as walls and trees erected on the ground, but if you encounter irregularly shaped obstacles, such as deserted railway, it will be overwhelmed. In addition, the system can only detect, do not recognize, it does not know whether they are pedestrians or road signs.
SICK LMS-291 point cloud (right) vs. Velodyne LIDAR (left) </ em>
In order to win, Hall brothers directly developed a new type of lidar, they will be a 64-wire laser launcher installed in the 360-degree rotation of the joint. As a result, the entire system can capture a real 3D image.
Unfortunately, their car failed to win the race of the year, but the LIDAR developed by the Hall brothers drew the attention of other teams. In the third and fourth (final) competitions, Hall became a laser radar supplier. In the last race, five of the six finished cars used the Velodyne LIDAR.
Why is laser radar essential for autonomous vehicles?
Uber test car is still using a rotating lidar </ em>
After that, Velodyne naturally became the ruler of the laser radar market. After Google created a large number of participants in the Challenge that year, the Velodyne LIDAR was used after the auto-driving division was established, and other companies involved in the development of autopilot became buyers of Velodyne products.
Right now, most self-driving cars carry three types of sensors: cameras, radar and laser radar. However, each type of sensor has its own "spear" and "weakness."
For example, a camera captures high-resolution color photos, but it does not have the ability to range and is almost as useful in measuring the speed of distant objects.
Radar ranging and speed measurement capabilities are good, the cost in recent years are also constantly declining. "When the object is closer to the vehicle, the radar works really well," explains Craig Glennie, a radar specialist at the University of Houston. "But because it uses radio waves, there is not enough detail in the radar at long range."
Lidar is taking into account the advantages of the above two sensors. Like radar, lidar ranging ability is very powerful, and some laser radar also has speed capability. At the same time, its resolution is also higher than the radar, which means it can detect smaller objects, and distinguish these objects are pedestrians, motorcycles or garbage dumps.
The difference with the camera, the laser radar in any light conditions are good.
Of course, the laser radar is not perfect, it is the biggest slot point is an expensive price. The Velodyne 64-line Lidar used on the car was as high as $ 75,000. Until nearly a year or two ago, it developed smaller and cheaper 32- and 16-wire products, but with a minimum price of $ 7,999. Late last year, Velodyne also introduced the new 128-line Lidar, which, despite its unusually powerful performance, is mulling the price of the Lidar.
Last year, Velodyne announced that it had taken a major contract with Ford to develop a new type of solid-state lidar which, once mass-produced, was able to drop below $ 500. However, Velodyne did not disclose the Ford detailed purchase volume and purchase price, as the solid-state lidar when mass production, it is still unknown.
Now Velodyne can not stop at any step, because the laser radar market competition is too intense.
The Rise of Solid State Lidar
For automotive manufacturers, of course, the harder the components are, the better, so the LIDAR with a wide range of moving parts is certainly not perfect in terms of reliability and price, such as the Velodyne rotary lidar.
For this reason, many experts believe that the laser radar to enter the mainstream market must be switched to a solid-state design, but this requires new devices to launch lasers in different directions to cover the vehicle surroundings.
Right now, the researchers come up with three main solutions for solid-state lidar:
The first is based on microelectromechanical systems (MEMS). The entire system requires only a small reflector to guide the fixed laser beam in different directions. Since the mirror is small, its moment of inertia is not large enough to move fast enough to track 2D scanning in less than a second.
MEMS LIDAR is now mainly developed by Luminar and Innoviz two companies. In addition, Lidar, called Infineon, recently acquired Innoluce, a startup that manufactures MEMS Lidar. In October last year, the domestic laser radar manufacturer Sagitar Polytron announced the latest development of solid-state lidar MEMS (Micro Electro Mechanical Systems) and OPA (Phased Array) solid-state lidar.
A big advantage of MEMS LIDARs is that the sensors can dynamically adjust their scanning modes to focus on specific objects, capture and recognize the details of the finer and smaller objects, which traditional laser lidar could not.
The second laser radar phased array design, it is equipped with a row of launchers can adjust the relative phase of the signal to change the laser beam emission direction. If the transmitter emits the laser simultaneously, the laser will fire in the same direction. However, if the left transmitter phase is on the right side, the laser will fire to the left and fire to the right.
Phased array LiDAR working principle </ em>
Although the principle is straightforward, most phased array lidars are still in the lab. Alex Lidow, CEO of Efficient Power Converter, said: "I have to say that the Phased Array Lidar is the future, and now we are still in the era of rotary or MEMS Lidars. The former is even more dominating."
When it comes to phased array lidars, Quanergy has to be mentioned, the startup, the key technology consultant for Strobe, and last October Strobe was taken over by GM, so GM might be tackling the technology.
The third is Flash LiDAR </ strong>, which runs more like a camera. The laser beam is diffused directly in all directions, so the entire scene can be illuminated with just one flash. The system then uses a miniature sensor array to pick up laser beams reflected in different directions.
One of the great advantages of Flash LiDAR is its ability to quickly document the entire scene without the hassle of moving targets or Lidar during scanning. However, this approach also has its own flaws.
"The bigger the pixel, the more signals you have to process." Pushing a large number of pixels into a photodetector is bound to cause various disturbances, with the result of a reduction in accuracy, "explained Sanjiv Singh, a robotics expert at Carnegie Mellon University.
Of course, if you have more interest in solid-state lidar, on January 16, 2018, Lei Feng network will drive in GA Silicon Valley smart driving summit held in Silicon Valley, we invited heavy guests such as the world's top laser radar company Velodyne CTO Anand Gopalan gave a keynote speech. He is the "key player" behind Velodyne solid-state and ASIC development. More summit information Click https://gair.leiphone.com/gair/gairsv2018, so stay tuned.
Detection range is a big limitation of the vehicle lidar
In short, the Flash LiDAR has no "hyperopia" and is not suitable for remote sensing in practice, and industry experts are convinced that LIDARs on fully autonomous vehicles have to see objects 200 to 300 meters away at a glance.
Jim McBride, now a Ford executive, also organized a team to participate in DARPA's autopilot contest in 2005, explaining in an interview why the experts made that judgment.
McBride first hypothesized a scenario where autonomous vehicles want to import traffic at high speeds. "At the time the traffic was at least 100 km / h, which meant that the car could run 30 meters per second, in which case most cars took at least 6-10 seconds to lift to 100 km / h, so they had to At this time you can see the vehicles around you, that is 180-300 meters away from their own vehicles. "
For any Lidar system, it is not easy to fire a laser at 300 meters and probe its reflected signal. The major manufacturers also resorted to the power of feeding to find some ways to increase the lidar detection range.
Today, most laser radar sensors are in the near-infrared laser range, the average manufacturer will choose this wavelength of 905 nm. However, it is closer to the wavelength of visible light (red wavelength is about 780 nm), while the laser will damage the human eye and burn the photosensitive detection cells on the retina, so the power of 905 nm laser is severely restricted.
In order to avoid harming the human eye, developers decided to switch to another wavelength laser. For example, Luminar has developed a lidar that uses 1550 nm lasers. It is much safer for the human eye because it greatly exceeds the range of visible light.
Solve the security issue, the power of 1550 nm laser radar can be greatly improved. Relevant data show that researchers have increased the lidar power 40 times, it is much more difficult to detect remote laser signals. However, there are bound to be lost, 1550 nm lasers and detectors are expensive because they require more specialized materials.
In addition to boosting the laser power, we can extend the lidar detection range by enhancing the detector's sensitivity. Last year, Argo AI, which was paid off by Ford Highgate, acquired Princeton Lightwave, a company that makes lidars with high-sensitivity detectors (single-photon avalanche diodes). This detector is quite sensitive, suitable for a photon in the frequency can be activated.
At the same time, these highly sensitive detectors are also new and have been used in military and exploration fields for many years. Last year, Princeton said it would introduce the technology to the automotive market.
TOF vs CW FM
In the development of a new generation of lidar, manufacturers are also faced with a third design new choice, that is, how to measure the time and distance. Most LIDARs use straightforward time-of-flight methods. It emits a very short pulse and then uses super-precise timers to figure out how long it takes for the pulse to go back and forth.
In addition to measuring distance by flight time, some vendors have developed more sophisticated methods and named it CW Wave Frequency Modulation (CWFM). As the name suggests, this method emits a continuous laser beam to the target that splits into two beams, one that fly toward the target and reflect back and the other reconstruct the beam during the flight.
The laser beam emitted from the laser radar will steadily increase in frequency, while the two beams separated by it will fly a different distance and will have different frequencies after recombination. This creates an interferogram with a beat frequency, from which researchers can derive the flight distance of the first beam.
There are several advantages to this approach. First of all, "Continuous wave frequency modulated lidar is not afraid of the background light at all," explained Paul Suni of the U.S. arms giant Lockheed Martin. "Conventional time-of-flight LIDARs may be overwhelmed if they encounter other light sources of the same frequency." At the same time, the CWFM system is more sensitive and works well even with intense glare.
This is very important because in the future, autonomous driving vehicles generally carry multiple lidar sensors, a large number of lasers fly in the air and the traditional time-of-flight lidar is sure to be disturbed.
There is also a big advantage of CWFM, that is, you can find the distance and speed of the detection object. In this regard, Suni explained: "If your sensor and the road between the relative movement of vehicles, then the signal will be Doppler shift.If you only measure one frequency, it will be a problem." In contrast, The CWFM Lidar determines the two frequencies that are constantly increasing and decreasing. By calculating, distance and speed problems can be solved.
For the moment, a start-up called Aeda, which deals with CWFM lidar technology, also reports specifically in The New York Times. In addition, Strobe mentioned above may also be undergoing similar research and development.
Low-cost laser radar on the road?
Indium Phosphide Wafer </ em>
Lectures on how to design laser radar in the end, I am afraid that various companies have tried various possible solutions, but even experts are not assertive in the end what kind of design can be the last laugh. However, everyone is confident that the next few years, the laser radar will usher in a big price reduction.
Numerous times of history tells us that the original impatient products will eventually fly into the homes of ordinary people, provided that mass production.
Lidow thinks the lidar will repeat the popularity of the ABS. "In 1979, I was developing ABS at GM," Lidow said. "At the time, the system was down to $ 8,000 and it was too expensive for the consumer market, but airlines could afford it, and even big trucks enjoyed the new technology as prices for the ABS system went down. But the truck has 18 wheels.
A few decades later, a set of ABS hardware prices only 20 US dollars.
"I've witnessed the evolution of this technology for myself," added Lidow. "Car companies can not wait to save on unnecessary costs, and the Lidar system is no more complicated than the ABS system of the year."
Lidow boldly predicted that the future price of laser radar sensors may drop to 10 US dollars level.
Singh of Carnegie Mellon also believes that the big drop in laser radar is at hand. "The Calculator was just about to be sold for $ 1,000 a day, and as output went up, prices started to fall and people were putting together a variety of components on a single chip."
The story is happening in the laser radar industry as researchers try to stuff all the sensor components into a single chip.
"Our laser radar chips come from 300-mm wafers, and their production costs can drop to as little as $ 10 a year if they reach the million-year mark," MIT researchers Chris Poulton and Michael Watts wrote in a paper. This chip uses optical phased array technology, does not require mechanical parts.
Of course, the goal of producing millions of dollars a year is hard to come by. Some analysts even warned that such products may have to wait in the lab for several years before they can really enter the consumer market. "All companies are trying to produce low-priced, solid-state products that will replace Velodyne LIDARs, but I'm probably not that easy."
In addition to the mass production challenges, solid-state lidar has a huge disadvantage that is its rather limited field of view.
Ford's McBride said phased array system "can only see 50 degrees of the scene", and now no one can break through this bottleneck, and the MEMS field of view is only 30 to 60 degrees.
This means that, to replace the current top-mounted rotary lidars, at least 6 to 12 solid-state lidars should be installed in the car, the price advantage of solid-state products will disappear. Therefore, it is too early to conclude that a mechanical laser radar will die. "A rotating disc is less expensive, and do not forget, there are a lot of mechanical parts that spin in the car, and they do not matter if they use tens of thousands of kilometers," Lidow said.
However, the history of the wheel rolling forward, although want to drive down the price of laser radar engineers need continuous efforts, but the road there is no hardship to overcome, so high-quality laser radar prices fell to 1000 or even 100 US dollars are possible.
This means that the price of autonomous vehicles can eventually be accepted by the average consumer. Prior to this, a 75,000 US dollars laser radar prices enough to prohibitive.
Right now, a lot of capital is being injected into the LIDAR industry in order to develop LIDAR products that are more powerful and cheaper. Although we can not predict when the LIDAR is cheap now, it is conceivable that LIDAR, the so-called "favored" high-priced laser radar, will get cheaper over time, with performance Will be better and better.
Finally, if you have more interest in Lidar, on January 16, 2018, Leifeng Xinqidian will host the GAIR Silicon Valley Smart Driving Summit in Silicon Valley. We invited heavyweights such as Velodyne CTO Anand, the world's top laser radar company, Gopalan gave a keynote speech. He is the "key player" behind Velodyne solid-state and ASIC development. More summit information Click https://gair.leiphone.com/gair/gairsv2018, so stay tuned.
Lei Feng network Recommended reading:
Is Velodyne having a solid win in the ongoing LIDAR World War?