Kautex Clear Vision Systems and XenomatiX True-Solid-State LiDAR are currently in a development partnership to develop LiDAR solutions for adverse weather and LiDAR cover contamination. We sat down with XenomatiX CEO, Filip Geuens, to discuss the trends in the industry and how our companies can work together to provide answers to some of the toughest challenges in the market.
Tell me a little bit about some of the XenomatiX’ latest technologies.
That is our solution for making solid-state LiDAR much more compact, robust and cost effective. The latest evolution is, of course, an increase of the number of laser beams in a product we released earlier this year, XenoLidar-X, by a factor of three, and at the same time, we also reduced the power consumption by a factor of two. This is all on the road towards mass production. We are continuing with the evolution to make the LiDAR less power consuming and more compact.
That is one aspect. The second aspect is the addition of a detection layer. It’s not only about generating a point cloud; in the end nobody's interested in a point cloud. It's all about detecting objects. And we are not only a hardware company, we also have significant software activities, with software to interpret point clouds and to extract a list of detected objects. We do this based on what we call four-dimensional artificial intelligence. It's an A.I. based method which takes advantage of the rich data sets that comes from solid-state LiDAR.
What are some of the biggest trends that you're seeing in future LiDAR developments?
At AutoSens, this year, there were a few LiDAR presentations. It struck me that many of these presentations were about VCSEL-based LiDARs. One clear trend is that VCSELs are recognized as a meaningful laser source for LiDAR, much for the same reason that XenomatiX was created: developing a robust and compact LiDAR. Since VCSELs are a chip and semiconductor components, they are very compact. In the same regard, I can say that the multi beam concept, which XenomatiX pioneered, is also adopted by some other companies. To answer your question shortly: whereas technology was widely explored a few years ago, we have now entered the phase of consolidation. It is becoming clearer which technologies will be the winning ones. And I can say multi-beam VCSEL based LiDARs are clearly becoming a trend.
What are some of the biggest trends you're seeing in ADAS in terms of number and size of sensors?
Car manufacturers are giving more attention now to level three, with some even talking about level two plus. This more gradual approach means that the car manufacturers are building experience by integrating perhaps just one LiDAR - typically a front view LiDAR - for highway pilots or a traffic jam assistance application. That’s very useful, because in this way the OEMs familiarize themselves with LiDAR technology, integration, and also with the sensor fusion aspect. One way of making the solution extremely reliable is by combining different sensor technologies and by double checking that what one sensor says is confirmed by another sensor. If there is a conflict, then it means you know it's a difficult case, a corner case [outside of normal parameters]. And in this way, a sensor fusion level three and a more gradual approach are clearly the trends on the ADAS side.
How important is communication within the vehicle and where do you see communication as a vital part of ADAS?
We think that not every sensor will have its own intelligence, object detection capability, and processing power. That would make it very expensive. This is what we would call a distributed architecture. We also believe that it's more meaningful to have a centralized architecture, meaning central processing power. And if the processing is done centrally, then communication becomes more important. It means that the raw data of all the different sensors is to be transferred to this central processor, which is a lot of data that must flow.
Our answer to this is to bring a LiDAR into the market which makes use of the communication network that is already available in a car: the communication network for cameras. For LiDAR we also make use of the CMOS detector, the output of which is basically an image. That image can be transferred in the same way as any other camera image. This is beneficial because it means we do not have to set up a special network for the LiDAR; we can just tap into, say, the camera network.
So, yes, communication is extremely important because of the trend towards centralized architecture. But let's simplify the setup and use LiDARs that do not have a specific need for a specific network, but that can use and reuse the existing network for other sensors.
What are some of the biggest challenges within ADAS and LiDAR development that you're seeing?
Meeting the cost target is extremely difficult, especially when comparing this with older technologies like camera and radars. How can you expect new technology to hit the same price targets right from day one? Again, that's an expectation that seems to be around in the market. If you look at, for example, navigation systems, in the beginning they were very expensive compared to what you pay today. And that's typically how new technology is introduced. LiDAR does not seem to get that possibility.
Another challenge is the fact that everybody hopes that LiDAR solves all the corner cases that cannot be solved by camera and radar. These technologies have been very valuable and there have been attempts to reach higher levels of automation with a camera and radar, which I can appreciate, but I think we are all in agreement that LiDAR has to be added. But then, of course, the expectation is that all the difficult cases that cannot be done with the technologies are now possible with LiDAR. It's only a matter of time until fusion of these three sensor technologies does indeed crack the nut and solves all the corner cases.
In terms of functionality, how does adverse weather impact LiDAR and sensor availability?
LiDARs are like cameras; they are optical systems, so they need a clear view, which is the laws of physics. A clear view also means that rain or rain drops, acting as a lens, should not influence the image. And so, yes, we must be aware of these limitations and must deal with it. That's why it's so important we work together with Kautex on ensuring a clear view.
“LIDARS ARE LIKE CAMERAS; THEY ARE OPTICAL SYSTEMS, SO THEY NEED A CLEAR VIEW.”
From the beginning XenomatiX focus was not just on the technology itself, but also on integration. Integration means finding a proper location for the LiDAR that is aesthetically pleasing, but it also means making sure that cleaning infrastructure is available to offset bad weather or adverse weather. Bad weather is not only a precipitation that is in the air, but it's even more importantly the precipitation and contamination that sits on the LiDAR cover.
How does XenomatiX overcome the challenge of adverse weather?
Part of the answer is by working together with Kautex, it's clear that this is why our technologies have to be brought together. The Kautex Allegro system works to clean LiDARs efficiently, quietly, and in all weather conditions. It uses both water and air to remove rain droplets, debris, and other contamination from the LiDAR cover.
Secondly, on the LiDAR side, we have taken some technical measures. The smaller the laser beams, the less likely it is the laser light interferes with something in the air like raindrops. Also, we simultaneously project many laser beams, which means we are projecting kind of a pattern. That pattern contains intelligence that is used to detect anomalies. If a raindrop acts as a lens and deflects a laser beam, we can detect it.
Thirdly, we have a detector with many, many pixels, more pixels than we have laser beams. It means we have pixels we can use to check the presence of contamination on the cover. And so we have made the LIDAR a little bit smarter to indeed also deal with these challenging situations.
Tell me a little bit more about the the joint development work between CVS and XenomatiX.
Sure. We have been collaborating for a few months only and a lot has been achieved. As I explained before, there are very good reasons why our know-how and technology must come together. If there is good synergy, then it's, of course, easier to get good results.
The collaboration is going very well thanks to transparency and with a good exchange of information. We also have seen new insights have been created on both sides. This is beneficial for both parties.
Finally, what we do together and what we achieve together is meaningful for the market. And it's not just a partnership for marketing reasons; it makes it more feasible and efficient to have well integrated LiDAR vehicles if we combine our technologies.
Do you see a need for an intelligent cleaning system?
Absolutely. A LiDAR is a visual, optical system. It needs a clean and clear view. And contamination on the car is important too; if the LiDAR is located high on the car, then it's a little less challenging, but when the LiDAR is in the lower position more contamination will hit the cover and then cleaning becomes essential.
Something else I want to add here is that, of course, Kautex has been doing cleaning for many years, but so far, the cleaning was mainly to help the human driver. And if the cleaning is not perfect, the intelligence of the human driver will fill the gap. But now the LiDAR and the other sensors are supposed to kick out the human driver in terms of the driving process, so the cleaning becomes even more critical.
The bar is also raised for fast cleaning and cleaning material. The cleaning liquid cannot stay on the cover for too long, or else it may blind the sensor. So the way cleaning is accomplished is becoming more challenging and complex. We must do it better, faster, with less impact on the clear view.
“AS CLEANING IS BECOMING MORE SAFETY-CRITICAL, WE MUST DO IT BETTER, FASTER, WITH LESS IMPACT ON THE CLEAR VIEW.”
