Radar sensing has been used for various functions in passenger vehicles for more than two decades. Like all sensors intended an automated vehicle (AV), radar has strengths and weaknesses. It’s good at detecting objects from long distances of a couple hundred meters or more, even in bad weather. And it has a low cost, which depending on quality and scale of production, is between about $50 to a few hundred dollars per unit.
But the relatively inexpensive, planar, phased-array radar antennas commonly used for functions like adaptive cruise control don’t have the resolution needed for higher levels of automated functions. That shortcoming motivated Hao and John Xin, the brother co-founders of Lunewave, to commercialize a spherical radar antenna based on the Luneburg lens design created three-quarters of a century ago. Tucson, Arizona-based Lunewave was founded in 2017.
German-American Rudolf Luneburg, a professor of mathematics and optics, first proposed his unique gradient-index lens in 1944. The lens, which serves as an efficient passive reflector, is used in fighter planes to enhance the signature of a radar signal. But because Luneburg lenses are commonly between the size of a large grapefruit and a soccer ball, they are too big for use in cars. That’s unfortunate, because the intricate pattern of chambers and branches inside a spherical Luneburg radar antenna imparts 360-degree sensing capabilities. The design also allows for fast, single-snapshot readings rather than comparatively slower conventional radar scanning.
Downsizing via 3D printing
Nearly 15 years ago, Hao Xin, a professor of electrical and computer engineering at the University of Arizona, started exploring how Luneburg’s approach might be used in new applications, such as autonomous driving and 5G communications. Xin, who earned a PhD. in physics from the Massachusetts Institute of Technology, now heads the University of Arizona’s Millimeter Wave Circuits and Antennas Laboratory. He’s also the chief technology officer for Lunewave.
Xin had been working with 3D-printing, polymer-jetting techniques since 2007. After two years, he was able to demonstrate the ability to “print” the Luneburg-inspired sphere, which contains 6,500 uniquely designed miniature chambers. He reduces the Luneburg antenna down to the size of a ping-pong ball – while maintaining the necessary complexity of the dielectric properties created via the thousands of tiny chambers. Lunenwave’s antenna can focus a returning radar signal to one focal point or scale that up to hundreds of points of focus. The current version, produced with an acrylic, weighs about 10 grams.
The lens-like antenna is placed inside an enclosure that contains micro-electromechanics to emit a radar signal. “The transmitter sends out the radar wave,” explained John Xin, Lunewave’s chief executive. The returning signal enters the sphere, which generates a high-gain beam in the opposite direction and is received by tiny receptors. “We’re not sequentially scanning,” said Xin. “We’re literally taking a snapshot of all around the 360-degree field of view.”
Current radars electronically sweep the beam across a view, explained Zachary Omohundro, senior scientist at Motivo Engineering, a contract design and engineering firm with facilities in Gardena and Fremont, California. “If Lunewave has a system that’s more like taking a flash picture where you get all of the data back, and you get all the data as one snapshot, that has interest,” said Omohundro.
Omohundro explained that a sweeping beam creates the equivalent of motion blur that can distort the measurements of a vehicle. “More than improving processing-time latency, it’s a matter of getting data from the world at the same time instead of having it smeared across time as you get different data points.” As with other radars, Lunewave’s long wavelength can penetrate particles (even some solid objects), so it’s not affected by bad weather, even a blinding snowstorm that can defeat lidar.
Because the signal is emitted in all directions and the antenna is a sphere, a Lunewave radar device – positioned on a vehicle’s roof – could provide 360-degree point-of-view data for objects as far away as 300 meters. Or if four such devices are strategically located at a vehicle’s corners, they could do the job required by a dozen or more traditional radars. Processing data from fewer radars also makes for more-efficient sensor fusion. But Omohundro cautions that unless there’s a high roof-mounted position, a radar with 360-degree capability might encounter blind spots based on the vehicle’s geometry. “How do you prevent the vehicle from occluding the view?” he asked.
Nonetheless, John Xin says conventional radar has “tunnel vision.” According to Xin, those traditional radar sensors with 2-degree angular resolution run into problems with objects at various distances. “As soon as you move outside 10 to 15 degrees, the distance gets significantly shorter, and the clarity gets worse,” he said. The angular resolution for Lunewave’s radar is 0.5 degrees, a field of view that approaches the capability of some lidars.
Powerful little Death Star
Critically, Lunewave also can modify the field of view based on each automaker’s requirements. “The design on the field of view, whether it’s the azimuth or the vertical, is dependent on the mounting location,” John Xin said. “For more-futuristic L4 or L5 [vehicle automation], folks are probably more inclined to put something on the roof rack, but for the L2 vehicles out there today, putting it in the front bumper works. It’s solely dependent on customer demands.” Xin said the field of view could be modified both horizontally and vertically. “If someone wants 180 degrees on the front bumper, and 90 degrees of coverage vertically, we can do that simply by putting additional receptors behind the sphere.”
Lunewave claims its system allows data to be calculated 10 times faster than traditional methods. It also allows the device to avoid interference, which can completely shut down a radar signal in the crowded 76- to 81- gigahertz spectrum. The size and design of the little round ball, which bears some resemblance to Star Wars’ Death Star, also can be modified for different bandwidths. John Xin admitted that there’s nothing unique about the company’s radar-circuit chipset. “The difference is the software algorithms and the little Death Star,” he said.
Hao Xin has extensive experience with 3D printing. “We evaluated more than 20 different 3D printers and more than 100 different materials before we selected the very special 3D printer that we use and our own process,” he said. The novel 3D-printing strategy might face production challenges, according to Omohundro. “I think they're going to be hard-pressed to be cost-competitive,” he said. “Most of the componentry that goes into today’s radar is all supported by existing chip-fab manufacturing processes. So it's hard to beat the cost-scaling that happens with that kind of silicon fabrication.”
John Xin insists that 3D printing the small antenna with commercial machines is scalable to the thousands per day. For mass production, Lunewave intends to partner with a Tier 1 supplier. Discussions with suppliers for volume production by mid-2020 are underway, as are pre-development projects with automakers.
In September 2018, Lunewave announced that it raised $5 million in seed funding from Fraser McCombs Capital, BMW i Ventures and Baidu Ventures. The company is currently finalizing a partnership that will allow it to ramp up production from 100 units a day to thousands. John Xin said Lunewave is now in the process of closing its next round of funding.Continue reading »