The big challenge is to come up with a LiDAR design which hits the performance and robustness targets and is able to be produced in mass market quantities at mass-market costs. Existing classical LiDAR designs, like mechanically spinning devices, most likely will not be able to achieve these targets, as many efforts in the last years have proven, says Florian Petit, Founder, Blickfeld GmbH in an exclusive interview with Geospatial World.
Top view of a street scene taken by long range Blickfeld LiDAR. The street coloured blue and trees in orange. 200 scan lines, about 120m, Full daylight.
Tell us about your company Blickfeld and its vision?
Blickfeld was founded in 2017 in Munich and develops proprietary, cutting-edge LiDAR-technology based on silicon MEMS mirrors and off-the-shelf components. The Blickfeld LiDAR meets the highest performance requirements at the cost and size needed for mass-market adoption. In 2019, Blickfeld will bring its first serial product – the Cube – to the market. It is designed for autonomous navigation, HD mapping, and other LiDAR applications.
Could you please elaborate on MEMS LiDAR technology and the advantages it has over existing LiDARs?
Alternative beam deflection units are MEMS (micro-electro-mechanical system) mirrors. These silicon components are produced in photo-lithographic etching processes. The advantage of MEMS technology is its precision down to nanometer level and exceptional scalability. Basically, hundreds of MEMS mirrors can be obtained from standard sized silicon wafers of e.g. 200mm diameter – all of which are produced in a highly accurate manner. Mass production with the highest precision is thereby enabled.
Today, the most widespread types are spinning LiDARs. They are often mechanically complex, resulting in large and heavy devices that are effortful and expensive to produce. The mechanical complexity stems mostly from the beam deflection unit, which deflects the optical beams (laser & reflected light). These spinning parts like gears and motors are subject to friction and abrasion. That is especially inconvenient because precision requirements for LiDAR are very high (typical LiDAR applications require an angle accuracy < 0,5°). The devices are prone to wear or even failure resulting in the need for regular maintenance. This problem with spinning systems has led to the demand for solid-state LiDARs where a robust and durable operation is expected.
MEMS mirrors have been around for several years and even find application in everyday devices like video projectors. The problem with many of today’s MEMS mirrors is that it is hardly possible to use them for long range scanning LiDARs. Long range performance and a wide field of view are important KPIs of LiDARs. Typical MEMS mirrors have small mirror sizes and small deflection angles, and as a result, the desired range and field-of-view performance can’t be achieved.
To overcome these limitations, Blickfeld invented a special LiDAR MEMS. It is specifically designed for the LiDAR application and optimized for maximum range and field-of-view. It has a mirror size of >10mm. Thereby, a maximum degree of incoming light can be directed to the photo-detector, ensuring the long-range capability of 150m and more on low reflective targets. Further, the MEMS mirror can achieve mechanical deflection angles of up to 120° which results in a wide field-of-view. It is produced with standard silicon processes which are cost effective and enables highly automated production.
Another huge advantage of the LiDAR MEMS mirrors is that the resulting optimized LiDAR design is quite simple: Besides of a few standard optical components, there is one laser, one 2D mirror unit and one detector. With these components, a packaging size of just 80x60x50mm can be achieved. The laser fires into the mirror, which deflects the light, providing the motion that conventional LiDARs achieve by spinning. The incoming reflected light is guided on a similar path and finally hits the detector. This so-called coaxial setup with identical incoming/outgoing optical path has the advantage of a very high sunlight suppression resulting in high SNRs.
Apart from autonomous vehicles, are there any other applications of MEMS technology as well?
MEMS is already used in various applications, such as inkjet printers, displays, accelerometers in cars (airbag deployment for example). Taking MEMS and using it for LiDAR applications is a quite new concept on the other hand. What makes it difficult is that MEMS is all about reducing size. That’s contrary to what you want in an optical sensor, because a larger size means capturing more light and therefore seeing further. What we did was design a MEMS mirror specific to these requirements, resulting in an unusually large aperture for MEMS, which means long detection range and a wide field of view, while still benefiting from the advantages of MEMS (scalability & size).
How do you foresee the future of solid-state LiDARs, and is there any cutting-edge innovation that has the potential to disrupt the market anytime soon?
Solid-state LiDARs have a bright future. They`re perfect for autonomous applications, as they are robust and less expensive and smaller – allowing a more seamless integration. There are several ideas as to how to make LiDARs solid-state. But we think the real market disruption will be producing solid-state LiDARs at a compatible cost and consistently high performance in large masses – and we believe that will be achieved with MEMS technology.
Where do you think the LiDAR industry is heading, and would the same trajectory continue?
Most experts agree that LiDARs are an integral part of every autonomous car in the near future. Studies assume that by 2035 there will be a total of 54 million autonomous cars on the roads. In 2017 the market size for LiDAR applications was estimated at $1.69 billion. It is projected to grow to $3.455 billion by 2023.
Regarding cost, size, precision and robustness requirements, solid-state LiDARs are considered more adequate than conventional, mechanical LiDAR systems. We believe LiDARs will become a mass-market product that enters all different areas of modern life where machines need to navigate in their environment.
What do you think should be done for the mainstreaming and easy availability of LiDAR sensors?
In our opinion, the solution is not to simply buy expensive components and build expensive LiDARs with them, in the hope that they will eventually get cheaper if produced in masses. Also, using technologies which reportedly will require decades of research will not lead to a timely solution.
In our opinion, many excellent components and technologies exist which allow to build a suitable LiDAR today. Appropriate components are in heavy use and well tested, with multi-million devices deployed in the markets. Examples include automotive grade lasers (905nm laser diodes), detectors with a resolution so high they can detect single photons (SiPMs, silicon photon multipliers) or silicon-based micro mirrors (MEMS mirrors). However, often they are not yet optimized for the LiDAR application – or the other way around, no LiDAR design has been found which is well tuned for these components.
So, we think the key is to adapt, tune and optimize each component and the overall LiDAR design towards each other. So, to tune optical and electrical properties of the laser/detector to the emitting/receiving optics and the beam deflection unit. To use an overall LiDAR design which suppresses disturbing light (sunlight) as much as possible (co-axial design). And – probably most importantly – to optimize the beam deflection unit regarding range, field of view and environmental conditions. This task is very complicated, as the components in themselves are complicated, LiDAR design is complicated and there are further boundary conditions like the need for highly automated producibility.
This constitutes a great challenge for the development team, as experts from various disciplines like optics, MEMS, signal processing, electronics, software and production technology have to work hand-in-hand.
We at Blickfeld identified the complexity of mechanical beam deflection units as a main weak point of LiDARs. This weakness affects both performance and production scalability. Therefore, we invented a special silicon-based beam deflection unit (MEMS mirrors) specifically designed for the LiDAR application. By using it, it is possible to build a performant and production scalable LiDAR module. This helps us in overcoming many of the problems of spinning LiDARs.
Cost is a factor that deters many from adopting LiDAR technology. How do you think would costs go down significantly?
By using silicon MEMS mirror technology and off-the-shelf components, costs can be cut down to a fraction of the cost of today’s spinning LiDARs when produced in masses.
For an average joe consumer, what are the various benefits that LiDAR can offer?
LiDAR is a sensing technology for high-definition environment perception. LiDARs can provide highly accurate distance images of the environment. The captured 3D data directly represents basic geometric properties of the scene like shape, size and distance. In mobility applications, where it is about motion and thus covering a certain distance in a certain time, this 3D data is essential to safely perceive the surrounding and to operate and navigate in it. Therefore, most perception sensors in cars, be it ultrasonic, radar or even 2D cameras with image processing, aim at gathering 3D information. LiDARs are especially well suited for this task as they can directly provide high resolution and accurate 3D point clouds in a quality unrivaled by any of the other sensor technologies.
LiDAR is a key sensing technology that finds application in various industries. LiDAR is an enabler in many areas and will drive the development of autonomous driving, smart city applications, robotics, drones, and many more. Most average customers will experience the results of a wider LiDAR deployment in an indirect way. They will not be stuck in traffic, they will be able to take autonomous cabs from a to b, and their packages will arrive faster thanks to optimized logistic chains.
Intel-owned Mobileye has developed a new autonomous driving system that makes use of high-definition cameras and not LiDARs. Would it have any impact on the automotive LiDAR industry?
Cameras serve different uses in autonomous cars than LiDARs. Cameras detect 2D colour images, needed for “seeing” traffic lights or road signs. However, when it comes to measuring distances or velocity, cameras have major drawbacks. To get that kind of information from a camera image, you require complex algorithms, and even then, there are limitations. LiDARs provide accurate, long and short-range, high-resolution 3D information about the surrounding environment and can be used for ADAS applications such as parking and traffic jam assistants, blind spot detection, lane keeping, autonomous emergency braking, and adaptive cruise control, as well as enable level 4 and 5 autonomous vehicles.
What are the major challenges ahead of the LiDAR industry?
LiDAR technology has been in use for decades and there are products available, which reach a high standard of performance. However today, LiDARs are often not robust, and very importantly – they are not widely available at a reasonable cost. That is a huge problem in the automotive market, since LiDARs find vast application in highly automated test cars today and are expected to be elementary parts of autonomous car sensor suits. Forecasts even see multiple sensors deployed in each car.
