To run safely, autonomous vehicles require a large amount of long range, high accuracy data regarding their environment. The cars need to analyze millions of points every second to build a 3D model of their environment. A widely accepted method to gather this data is to use a LiDAR sensor which incorporates a bank of lasers installed on the car, often the car's roof. Although LiDAR can use the laser bank to gather the required information, the sensor has traditionally been too expensive for commercial use.
Amidst the clash of goliath corporations over LiDAR tech, some companies are trying to lower the price of LiDAR for all.
LiDAR's Notoriously Expensive Implementation
Some LiDAR sensors are as expensive as the car itself. For example, the high-performance sensors employed by Google’s early self-driving cars were about $70,000. These exorbitant prices encourage other car manufacturers, such as Tesla, to choose a different path and use a combination of cameras, ultrasonic sensors, and radars rather than LiDAR units.
Raw data from the HDL-64E. Photo courtesy of Velodyne.
On the other hand, there has recently been a tremendous amount of research to lower the cost of LiDAR sensors while trying not to compromise on its performance. While LiDAR generally consists of a number of high-quality optical components inside a spinning housing, several companies are attempting to design solid-state LiDAR sensors which consolidate the bulky components into a single chip. In this way, they hope to reduce the cost significantly and pave the way for the commercialization of any LiDAR-related technology.
One of these companies is Velodyne LiDAR Inc., currently the global leader in the LiDAR industry. While the company already has a virtual monopoly on the LiDAR market, Ford and Chinese search giant, Baidu, have jointly invested $150 million in the company to make affordable LiDAR a reality.
Velodyne’s Solid-State LiDAR
Velodyne announced a new low-cost LiDAR which can revolutionize many industries such as autonomous vehicles, ridesharing, 3D mapping, and drones. The new sensor, developed in partnership with Efficient Power Conversion (EPC), has fewer moving parts and offers a more reliable operation than conventional sensors. It can also be easily integrated into a larger system. The company expects that, when mass produced, the solid state LiDAR will be under $50.
According to David Hall, founder and CEO of Velodyne LiDAR, the new sensor has three main advantages over previous designs: it is smaller, less expensive, and more reliable.
The miniaturized sensor is a monolithic gallium nitride (GaN) integrated circuit that consolidates all the required components into a 4mm square chip.
Velodyne's new 4mm LiDAR sensor compared to a U.S. quarter. Image courtesy of Velodyne.
According to Dr. Alex Lidow, CEO and co-founder of EPC Corporation, GaN technology can lead to higher image resolution while increasing the integration level and, consequently, reducing the product cost.
Unfortunately, the company still refuses to share any detailed information about the new sensor’s performance.
Will Non-Rotating LiDAR Replace Conventional LiDAR?
Velodyne, which has shrunk the sensor into a solid state chip, is able to dramatically reduce the product cost. However, as Anand Gopalan, vice president of Velodyne's R&D notes, there is a trade-off between cost and field of view. For example, while the high-performance LiDAR sensors are as expensive as $70,000, Velodyne currently has other sensors with shorter range and a narrower field of view which cost about a few thousands of dollars.
He adds that the new sensor compliments the existing ones. There are numerous applications and each application needs to meet certain needs. For an autonomous vehicle application, a 360-degree field of view is mandatory; however, there are some advanced driver assistance applications which need a much narrower field of view. Velodyne is developing an engine which can cover both these applications. An application which relies on a 360-degree field of view will take the engine and put it in a rotating solution; however, a less demanding application will utilize the engine in a non-rotating solution.
According to Gopalan, Velodyne’s products are trying to marry the transmitter and receiver so that the signal to noise ratio can be increased. Moreover, the company has a unique approach in detecting objects which are either very close to the sensor or very far away from it.
The Megafactory Advantage
In January 2017, Velodyne opened a LiDAR Megafactory as well as an R&D facility in San Jose, CA to stay ahead of the growing competition. The new factory will not only increase production, it is large enough to do sufficient in-house long range testing on new LiDAR technology.
Velodyne's new Megafactory. Image courtesy of Business Wire.
Production of existing sensors is already underway, with the goal of producing over one million sensors in 2018. Opening this factory is a step towards decreasing the large costs traditionally associated with LiDAR technology.
Attempts to Design Cheap LiDAR
In a previous post, we discussed some companies working toward more affordable LiDAR technology. Since then, other players have come into the arena.
Innoluce, in the Netherlands, resorts to a microelectromechanical mirror system to arrive at a 100-dollar solution which offers improved resolution and range.
MIT researchers have leveraged silicon photonics to shrink a LiDAR to a tiny chip. The fabrication process is CMOS-compatible and researchers believe that the product cost will be about only $10. Currently, the sensor range is just two meters but the research team claims that there is a clear path to reach a 100-meter range.
In addition to advancements made in LiDAR design, some companies, such as Oryx, are trying to find a more radical solution.
According to Rani Wellingstein, the co-founder of Oryx, radars generally detect objects at 150- or 200-meters, but they do not offer enough resolution. On the other hand, LiDAR provides great resolution at the cost of a range limited to generally about 60 meters which even reduces to nearly 30 meters when in direct sunlight. Oryx attempts to design a sensor significantly better than the existing solutions.
Oryx technology, an Israeli start-up, has recently announced a new sensor which uses long-wave infrared lasers to illuminate the road. Since this terahertz laser is invisible to human eye, in comparison with a LiDAR, the new sensor can radiate at higher power levels. Besides, unlike LiDARs and cameras, the new sensor is not blinded by fog or direct sunlight. This is due to the fact that we do not have much sunlight radiation at these frequencies and water drops only poorly absorb these waves.
Moreover, similar to radar, Oryx’s sensor treats the reflected signal as a wave rather than a particle. In this way, the sensor can take the Doppler Effect into account and calculate the velocity of objects around the car. A large number of microscopic rectifying nanoantennas are used to sense the reflected signal. According to David Ben-Bassat, Oryx’s other founder, these nanoantennas are the result of a six-year endeavor.
Oryx has not revealed any performance metrics in terms of the sensor’s field of view; however, since the company is attempting to reduce the number of moving parts we can expect that the sensor cannot offer a 360-degree field of view.
Wellingstein claims that the new sensor has a million times the sensitivity of traditional LiDAR.
Affordable LiDAR will significantly advance the capabilities of numerous products, and the company who can make the first affordable LiDAR sensor will have a definite advantage. We could expect that these LiDAR units will bring high-resolution gesture recognition to mobile phones, allow package delivery drones to avoid collisions, and perhaps, one day, LiDAR will be affordable enough to be used in robots’ fingers so that they can recognize and manipulate objects with more confidence than ever before.