Will automotive radar return to analog in the future?

Now is the time to return the radar platform to analog! Startup Matawave believes that "we still exist in the analog world, and so do cars." The company hopes to change the limitations of traditional radar through a high-performance analog radar platform ...

To increase real-world situational awareness in highly automated vehicles, many automakers are beginning to accept the need to deploy a variety of sensor types around every chassis. However, what they haven’t considered is the quality of these sensors. For example, how do today’s vision, LiDar, and radar sensors perform? What requirements do automotive sensors need to meet?

　　Metawave is a startup that was spun out of Xerox PARC Research Center earlier this year, but is confident that it can change the "traditional radar limitations" identified by the automotive industry. Currently, automotive radar cannot "see" distant objects, nor can it distinguish what it sees. Its processing speed is not fast enough to operate while driving on the highway.

　　In short, today’s automotive radars can’t necessarily see everything that cameras or lidars can see. Their only saving grace is that they can operate in all weather conditions.

　　Metawave, which was founded in January to commercialize metamaterial radars and antennas under an exclusive license from PARC, is touting its “full radar package” technology. Metawave plans to showcase the prototype at the 2018 Consumer Electronics Show in January.

　　Metamaterials are small, software-controlled engineered structures deployed on printed circuit boards (PCBs) that the company claims can steer electromagnetic beams in ways that were previously limited to more powerful and expensive military systems.

　　However, Metawave does not blame the problems of today's automotive sensors on radar chips - mainly designed by suppliers such as NXP, Infineon or Texas Instruments (TI). In fact, Metawave's full radar package is not limited to a specific radar chip. Instead, the startup believes that the problem lies in the beamforming technology in the radar sensor (including the antenna), which leads to problems in resolution and speed.

　　Regression simulation

　　Maha Achour, CEO of Matawave, believes that it is time for the industry to "return radar platforms to analog." She emphasized, "We still exist in the analog world, and so do cars. Therefore, Metawave plans to create an affordable, high-performance analog radar platform without the complexity and cost of military-grade operations."

　　Metawave's analog radar technology is based on an electronically steered antenna. It uses a single antenna with dual ports, one end connected to the transmitter (Tx) or receiver (Rx) link, and the other end connected to the microcontroller (MCU). The MCU defines and controls the antenna beam width and direction using a lookup table (LUT), allowing Metawave's analog radar to scan at microsecond speeds. (Source: Metawave)

　　Achour claims that Metawave has designed a new analog radar using a single antenna that can steer and shape the beam horizontally and vertically, and adjust the beam from a wider field of view to a very narrow cone angle - as low as 1 degree. "We can do it very quickly - scanning at microsecond speeds," Achour said.

　　But how does Metawave’s analog radar compare to the digital radars now widely used in vehicles?

　　Radar based on digital beamforming (DBF) technology requires an antenna array to focus the electromagnetic signal sent by the transmitter in a specific direction and redirect it in other directions. The receiver then captures the return signal from the object and processes it digitally to form an image of the scene.

　　To accomplish this, Achour explained, digital radar must “inject different phase delays into each antenna to make the beam converge in one direction and spread out in other directions.”

　　The drawback of DBF is the phase delay. The calculation requires complex and lengthy digital signal processing. Achour pointed out: "This intensive signal processing leads to extremely slow response speed (millisecond delay in steering the beam) and poor "collective" radiation pattern because the beam is steered away from the antenna line of sight (zero degree angle).

　　Digital radar sensors currently used in vehicles use digital beamforming technology and calculate phase delays (i.e. weights in the figure - wi) through complex and lengthy digital signal processing. The antenna has static radiation and relies on digital weights to form and steer the beam (Source: Metawave)

　　As a result, she said, “these conventional radars cannot see at wide angles at long distances because they cannot control the main lobe and side lobes well.”

Making decisions about distant objects

　　“One of the biggest issues I think architects have to address with autonomous vehicles is being able to make decisions about objects that are far away from the vehicle,” said Drue Freeman, an advisor and investor currently working with Metawave. Otherwise, the top speed of an automated vehicle will be limited, Freeman noted.

　　“Today’s radar solutions, even with the best digital beamforming technology, may be able to reliably see 200 meters in front of the car and detect that ‘something’ is there, but they can’t identify what it is,” Freeman said.

　　The reality is that DBF supports either high resolution or high signal-to-noise ratio (SNR), but not both.

　　Super material

　　Metawave says its goal is to provide high-performance radars similar to those used to track missiles, but without the cost, complexity, and power consumption required for military applications. Metawave's analog radar "emulates a phased array," like military antennas, Achour said. But the startup can do this without relying on phase shifters deployed by military applications because it uses its own metamaterials.

　　Metawave's metamaterial frequency adaptive steering technology (Source: Metawave)

　　“What’s exciting about Metawave is its metamaterial-based analog beamforming technology, which allows them to precisely steer radar beams, achieve faster operation and better SNR without sacrificing resolution,” Freeman admitted.

　　Bernard Casse, chief technology officer of Metawave, said that in addition to enabling "vision" and "speed" for radars and antennas, Metawave's analog radar will also bring "intelligence." Metawave has embedded an artificial intelligence (AI) engine in its analog radar.

　　Inside the AI ​​engine is a series of algorithms, Casse explained, "including, in addition to deep learning and decision-making algorithms, range-Doppler evaluation algorithms, clutter and interference suppression algorithms, object detection and tracking algorithms, and other proprietary electromagnetic and radar code."

　　What exactly can the AI ​​engine in the radar learn? Casse said: "It is highly dependent on the scenario." For example, if a car passes under a bridge, it will encounter many signal reflections. The AI ​​engine can classify and sort through various interferences and help the radar locate the objects it must see.

　　“The Metawave case is very interesting because in many cases, radar will be the first sensor to ‘see’ something on the road, and the AI ​​engine can be used to initially classify what it sees before sensor fusion processing,” Freeman said.

　　"Every sensor has its limitations, so it goes without saying there are many examples of radar failures," said Mike Demler, senior analyst at The Linley Group. "But there are more likely cases where the software failed to interpret the signal correctly."

　　He pointed out that "the worst case was the accident involving a Tesla self-driving car in Florida a few days ago, which was caused by the Tesla's self-driving system not detecting that a white truck crossed its lane. Automotive radar has always been a relatively cheap sensor, mainly used for simple ranging functions such as adaptive cruise control, and it is not designed for object recognition. Apparently, Metawave is working on developing technology using synthetic aperture radar (SAR), which will provide radar with object recognition capabilities."

　　Opening up new business models

　　Metawave CEO Achour is optimistic about the great prospects of AI in its radar applications. Once the radar begins to use its AI "brain" to collect data on the road and interpret the driving environment, Achour hopes that Metawave can provide usable data to the automotive industry. "We can provide code-based AI and algorithms with the results learned from radar operations and earn service fees from them."

　　According to multiple forecasts, in the future Level 4/Level 5 autonomous driving stage, the automotive industry will no longer rely on unit sales of cars, but will focus more on the mileage of each car. In this case, Achour pointed out that hardware companies must also change their business models. Providing intelligence collected by AI as a service brings new business opportunities to Metawave.

　　Will automotive analog radar replace lidar?

　　“If Metawave can get the cost down, it might be possible,” Demler said. But he is skeptical that Metawave’s radar can really surpass lidar’s resolution.

　　Freeman believes it is too early to predict. He explains, "Each sensor has its own strengths and weaknesses. What Metawave has done is address some of the weaknesses of radar, and I think it has done so in a way that makes it powerful enough and high enough resolution that automotive OEMs may be able to design a full-stack sensor system that does not require lidar at all." However, he adds, "The lidars currently in use are high quality and low cost and are still more effective in accomplishing this task."

　　Achour sees it slightly differently. “Initially, all sensors will be required to achieve full autonomy,” she said. “As AI engines mature and digital maps become more reliable and accurate, radars and cameras will be enough to achieve zero-accident autonomous driving, even without V2X communications.”

　　“Some people might think that lidar is absolutely necessary to achieve precise positioning,” she said. However, she said from her own experience that Metawave’s analog radar (called Warlord) supports powerful 3D imaging, and combined with digital maps, “will be sufficient to provide precise positioning. I expect this to be available in the mid-2020s to early 2030s.”

　　Development Challenges

　　Metawave also faced challenges in developing a full radar package. George Daniel, vice president of engineering at Metawave, pointed out that Metawave's radar solution is designed to operate in the 76-81GHz frequency band.

　　By authorizing use of the entire 76-81 GHz band, the FCC has created a contiguous block of spectrum for long-range vehicular radar.

　　This means that Metawave’s “metamaterials need to interact with discrete components” that were originally designed for early automotive radar systems, using integrated 24GHz radar sensor technology that operates in the lower frequency band.

　　US$7 million in first round financing

　　Currently, Metawave's core team consists of seven engineers, including management. In September this year, it received $7 million in first-round seed funding from Khosla Ventures, Motus Ventures, and Thyra Global Management.

　　So how much more money does the startup need? “Maybe another round of funding!” Achour said. She expressed confidence that “Metawave’s technology can solve the most fundamental problems.”

　　In addition to planning to showcase its automotive radar prototype at CES, Metawave also plans to showcase its smart beamforming antenna designed for 5G networks at the Mobile World Congress (MWC) in Barcelona, ​​Spain in February next year. Achour explained that the current MIMO architecture cannot support speeds 1,000 times higher than 4G in the upcoming 5G era, and Metawave's smart beamforming solution can direct energy to specific user devices to provide the bandwidth required to optimize the online experience.

　　Achour is no newcomer to the world of metamaterials. She was co-founder and CTO of metamaterials company Rayspan, which created promising metamaterial antennas for cell phones about a decade ago but ultimately failed to make it.

　　How is this possible? Achour said that Rayspan's business model is based on licensing, designing antennas and RF front-end modules and then licensing them to customers. However, licensing has never been a business model suitable for hardware solution startups, because "startups do not make profits fast enough to support the operating expenses that must be paid after a few quarters of product sales." She explained: "This is why Metawave works with third-party manufacturing partners to build full radar sensors."

　　So what’s the difference between the metamaterials used by Rayspan and Metawave? “Rayspan’s antennas are passive, meaning their radiation pattern is fixed,” Achour said. “Metawave’s antennas are active, and they are smarter because they have active elements on board that allow the antenna to control its beamforming and steering.”