Radar startups review 2019
Should you check the review as of 2018 please refer to https://medium.com/frontier-tech-review/radar-startups-review-as-of-august-2018-59a75a0e648f.
The radar sensor market has hardly provided us with much news this year. Most companies are still working on their products, stirring up the interest with press releases and rendered videos. We all know the long and heavy development cycle of microwave technologies, however, this situation has affected the investors, who have put only 26M in radar companies for 9 months of 2019, compared to 39M in 2018 and more than 310M earlier.
Cash flow can be a sign of the industry’s cooling down. Yet, there are 15 companies in this review, (way more than in 2017), most of their money came from 2017 and earlier, and the amount of money raised in 2019 is less than in 2018. Moreover, we now have a startup that has been closed down and it is Oryx, one of the most well-funded companies. Many companies did not show any progress in the automotive sphere.
We can use this intermission to see a bigger picture and name the notable trends in automotive radar’s market. No one is building just a radar these days, therefore I would like to point out three major directions in the industry: imaging radars, complementing technologies and driver monitoring.
Imaging radars (Arbe’s radar is a nice example) is the holy grail of the radar market because it can probably be an ultimate sensor to complement optical cameras and bury lidars as a self-driving car sensor. Imaging radars have such a huge number of channels so they can provide as detailed picture as lidars do, with bonus velocity data. The concern is that imaging radars may be still too expensive for today’s and future ADAS and currently, self-driving cars are not evolving fast enough that it could be a market for series imaging radar. See why Oryx has closed its operations.
Imaging radars are also falling into self-driving carmakers' favor due to the fact that sensor fusion systems demand the same data format as provided by lidars.
Basically there are two major formats that represent the environment. First, it is an image or a two-dimensional matrix with outlines of objects complemented by a depth map. The matrix can be merged with camera information thus creating a complex 3D environment. The second one is a 3D point cloud, pretty much the same as lidars provide us with. If the sensor uses one of these types of data representation, a sensor fusion system developer could just change the source of the data provided keeping the whole system intact. Imaging radars are capable to work with both approaches.
Last but not least, multi-channel chips for microwave sensors seem not to face the technical issues which lidars do on their way to the mass market.
Complementing radar technologies are not that complex. Building an imaging radar means one has to develop a multi-channel transceiver. This is expensive and time-consuming. When a company is focused on something less complex, it likely may shorten the development cycle and bring the product to the market faster. Nonetheless, we see such companies still need to build the whole product to show its benefits. The examples are Ghostwave with their interference-free signal modulation or Lunewave with a new antenna type. There is a feeling that these companies have started with the technology but lately were forced to build the whole radar.
One of the key benefits of radars as a technology is its long range — up to 300 meters and more. But what if you have a short-range radar, which range is limited to a dozen or several meters? Driver monitoring and gesture recognition could be a kind of solution though radars do not show their best in these applications. The idea is that a small radar is scanning the driver’s side of the cabin and sees any signs of distraction based on how the driver’s head, body and/or hands move. It can also recognize the gestures which the driver uses for changing the temperature or accepting a phone call. The issue is that affordable cameras can do the same and even more: recognize the driver’s emotions or sleepy eyes, which is not possible by radars.
Ainstein
USA, Laurence KS / China, Shanghai, founded 2015*
* — here and further as mentioned in Crunchbase and/or other sources
The company with Chinese roots builds its automotive radars using 77 GHz 40-nm transceiver from Calterah, a notable Chinese chip manufacturer. The radar is claimed to be an imaging one, but technically it has less TX/RX channels than other imaging radars do. The obvious advantage is that Ainstein has launched its manufacturing facility in 2019, showing its readiness for scaling the production. The company provides radars with its custom sensor fusion system demo based on NVIDIA hardware as well.
Arbe Robotics
Israel, Tel Aviv, founded 2015, 23M raised
Arbe is the company that continuously focused on developing an automotive imaging radar and has passed some milestones already. This year Arbe disclosed some technical details about what they build. The radar implements sort of MIMO technology forming more than 2000 virtual channels from 96 physical receiving and transmitting channels that is called Enhanced FMCW technology by Arbe. The company has confirmed that it assumes a 300 m range with 0.1–0.6 m accuracy, 0.1 m/s velocity accuracy and 1–2 degree angular resolution, both azimuth, and elevation. Due to the enormous amount of virtual channels, the radar should be capable of building a 4D (i.e. 3D coordinates + velocity) picture of the environment. The radar will have its proprietary tracking, classification and localization algorithm, but allow customers to implement their own post-processing.
Arbe Robotics seems to be ready for pilot projects, none of which has been announced so far.
Artsys360
Israel, Holon, founded 2013, 4M raised
The company is developing an idea of a 360-degree radar with no moving parts. The developer has not yet pleased us with any news concerning their automotive lineup, there is no news but about UAV detection. Nothing is disclosed about rounds they raised as well.
Echodyne
USA, Kirkland WA, founded 2014, 64M raised
Although there is no news regarding the automotive application, Echodyne definitely has had a good year. FCC has approved their Echoflight radar back in 2018, and EchoGuard in 2019. Echodyne has participated in the first FAA-approved true beyond-visual-line-of-sight domestic flight of an unmanned aircraft system under the small UAS rule in the USA. It shows Echodyne’s ability to ship radars that function well in real conditions, and the company should have some products that are 100 percent ready for pilot projects in the automotive sphere. The 20M Series C round raised from new and existing investors may help Echodyne with it.
Ghostwave
USA, Columbus OH, founded 2016
The idea Ghostwave is focused on is to eliminate a risk of mutual radar interference, which is counted as one of the major radar issues, by sending millions of random signals that eliminate the risk of interference. It may be a radar with random waveform modulation, which is considered a state-of-the-art solution. The company has raised an undisclosed round from Ohio’s Rev1 Ventures at the end of 2018. Prior to that, GhostWave had received a $2.5 million development contract from the Ohio Federal Research Network.
Lunewave
USA, Tucson AZ, founded 2017, 5.5M raised
Now Lunewave’s not only developing its unique Luneburg antenna for W-, K- and X-bands but an automotive radar sensor operating in 76–81 GHz with a 360-degree field of view and detection range of 300 meters. Last year the Urban-X accelerator alumni has raised 5M for getting its radar and antenna products closer to the market.
Metawave
USA, Palo Alto CA, founded 2017, 17M raised
Metawave’s working on antennas based on metamaterials which are capable of analog beamforming and being used for high-frequency signals such as automotive radars and 5G communications. Their automotive product is called WARLORD and declares more than 300m range detection for vehicles and 200m for pedestrians. The company has claimed that it had started tests back in 2018, but there are no results have been disclosed.
What we know is that Metawave is still good with naming, their new AI-related technology for augmenting perception and early warning systems like predictive analytics is called WRATH after Warlord Radar AI Thinking like Humans.
Neteera
Israel, Jerusalem, founded 2014, 8.5M raised
Neteera has built a radar-on-chip for high-res short-range radars suitable for health monitoring, but it is also applicable to vehicle driver monitoring. There is no information about any pilot project up to date.
Oculii
USA, Beavercreek OH, founded 2013, 21M raised
Like a year ago, Oculii is developing its 4D-radar that should be giving unprecedented point cloud density of X, Y, Z and velocity information for each point. Its advantage is that such a system is capable of direct pixel fusion with camera or LIDAR systems, so it may be easy to enhance some existing sensor fusion setup with Oculii’s radar. According to the company’s profile on the Texas Instruments design network, it may be utilizing TI’s latest AWR 1642 chip for at least one of the applications. The company claims it is ready for demoing its product.
Omniradar (Staal Technologies)
Netherlands, Eindhoven, founded 2017, 3.4M raised
Last year Omniradar did not provide us with any news regarding its automotive applications. The probable reason is that the company has switched to industrial applications, water/other substances level measuring to name a few, under a new name Staal Technologies. Its radar-on-chip is used in SenZ2, the affordable level meter being promoted in 2019.
Oryx
Israel, Petah Tikva, founded 2009, closed down 2019, 67M raised (27M spent)
Sad news came from Oryx this summer as the company has been closed down in August. The website is not responding and its CEO has claimed that 40M remaining in cash will be returned to its investors. In the beginning, Oryx has been developing a novel technology that could bring together the benefits of radar and lidar, turned to a lidar later, so it was called a lidar company. The CEO has explained that the reason for shutting the company down is the market which is not growing fast enough for a venture-backed company. The issue may also lie in the technology field where Oryx could not develop enough affordable and reliable sensors.
Steradian Semi
India, Bangalore, founded 2016, 75K+ raised
It is not much disclosed about Steradian’s radar, but at least it is said that the company has 24 GHz imaging radar prototypes, which is notable since a 24 GHz band physically does not comply with “imaging” requirements due to its bad distance resolution. But Steradian Semi also has 77 GHz products.
Uhnder
USA, Austin TX, founded 2015, 75M raised
Recently Uhnder came out with some news regarding their technology and, as mentioned in Microwave Journal, amount of money raised for it. Backed by 75M investment, Uhnder in a second became one of the richest radar startups so far.
Unlike popular frequency modulated continuous waveform (FMCW) approach, Uhnder uses phase modulated continuous waveform (PMCW) with digital code modulation (DCM). Their radar on chip (RoC) has 12TX/16RX configuration (as explained, 8 TXs are operating at the same time in MIMO mode) to form 192 virtual elements in the 77GHz band. Processing is done on the same chip by two CPUs and two DSPs. The claimed processing power is up to 20 Tera-Ops, a huge computation ability, but it is not clear how the estimation was done. It is probably the first commercial radar that implements an SDR (software defined radio) principle.
Uhnder should be performing evaluations with potential customers right now, more details to come late 2019/early 2020. It had been disclosed earlier that Uhnder would be a supplier of Tier-1 supplier Magna’s radar called ICON.
Vayyar
Israel, Yehud, founded 2011, 79M raised
Focused on providing a 3D radar image of an object. Late 2018 they started with Walabot Home (https://walabot.com/walabot-home), a home surveillance system that respects one’s privacy by using radar instead of a camera. It is made possible by using
Vayyar’s Radar on a Chip (ROC) has 48 transceivers at 76–81GHz which allows over 2,000 virtual channels. The chip also comprises an internal DSP for real-time signal processing. This single-chip solution is easily integrated into the existing automotive framework, reducing the overall cost and number of sensors needed for the vehicle.
The obvious automotive application for this technology is driver monitoring. The company mentioned it may be also used for collision avoidance, but it may be a tough challenge since requires much more range and velocity than classic monitoring. There is no information about existing automotive setups, but Vayyar probably should be testing it.
Wavesense
USA, Somerville MA, founded 2017, 3M raised
Wavesense is another example of how various radar implementations may be. It claims that radar should not just look ahead, but down. That means a GPR, ground-penetrating radar.
It scans the ground for centimeters or meters under the car thus forming a unique pattern of any piece of the road with centimeter accuracy. First, this data is put into a geo-referenced database, which forms a new kind of map. Then, a GPR equipped vehicle compares its realtime data, fingerprints, to a map to understand its location. This approach is known for being tested in military applications by MIT’s Lincoln Laboratory, but civil cars may benefit from it as well. Theoretically, a GPR can supply a car with high-accuracy navigation with a lack of GPS signal, road signs or lane markings by using only its own maps. As mentioned, Wavesense is ready to proof and currently searches for pilot projects.
Zendar
USA, Berkeley CA, founded 2017, 10,5M raised
YC-alumni company Zendar did not get much into the technology of their high-res radar and their stage. The amount of money raised increased by nearly 6 million, which is spent, among others, for developing a digital processing accelerator. It is much more interesting to know what Zendar does about RF part of their radar, though.
Don’t hesitate to contact me via koshurinov@gmail.com.