How Ground-Penetrating Radar Could Help Self-Driving Cars ‘See’ Better
This could be the next frontier in autonomous tech, but it has a lot of mapping to do.
First patented in the early 1900s, ground-penetrating radar has been used by geologists, archaeologists, and aeronautics engineers ever since. Notably, the Apollo 17 mission used a GPR to record depth information about the moon. GPR is not new technology, but it’s new to the application of advanced driver-assistance systems (ADAS).
Some companies are betting on GPR to get ahead of their competitors in the field by improving the reliability and accuracy of autonomous features. Within the hot ADAS segment, everyone is looking for the panacea that will make billions of dollars and solve all of the autonomous driving challenges.
“People are familiar with global positioning systems, which triangulate location within about a meter,” says IDTechEx technology analyst Dr. James Jeffs. “For navigation, that’s great, but it’s not good enough for ADAS and the level of autonomy that we expect.”
WaveSense is one of those on the edge of the GPR trail, using a radar system designed to penetrate the ground and take measurements at centimeter-level accuracy. Its application for GPR was spun out of the Massachusetts Institute of Technology's Lincoln Laboratory, where it was being developed to automate military vehicles. They quickly realized it filled the same gaps for commercial vehicles.
The company’s CEO and co-founder Tarik Bolat says precise localization in any type of dynamic environment (such as poorly-marked road surfaces, off-road terrain, and inclement weather) is going to change the direction of technology for ADAS. Ultimately, GPS, lidar, and cameras map the world around you and track to that, including buildings and other markings. Bolat says WaveSense takes that deeper, measuring routes, rocks, cavities, and soil about three meters into the ground.
Why image the subsurface? Because it’s stable over a long time period, Bolat says. The subsurface doesn’t change that much overall, which is critical for precise measurements.
ADAS uses various data inputs to enable safety features like automotive imaging, which extends the capabilities of the human eye for high visibility in inclement weather situations and low-lighting settings. Lidar adds more sensors and cameras, and in conjunction with GPS and Wi-Fi, late-model vehicles are not only trackable but tuned into their surroundings. Here’s the thing: when lane markings aren't visible or the road surface is obscured in some way, optical sensors typically fail. Mapping three meters down adds a level to the existing sensors and cameras; GPR is not meant to replace existing sensors but add another layer of data points for engineers to explore and build upon.
Essentially, if GPS is a snapshot in time, GPR is a fingerprint. Together, it creates a more robust moving picture.
WaveSense, in particular, seems to be making some inroads into the crowded ADAS tech market even if mass adoption isn't imminent. The company announced a $15 million round of funding earlier this year, and it was named best-in-show at the North American International Auto Show in Detroit in 2019. Then in 2020, former Ford President Joe Hinrichs was appointed to WaveSense’s board of directors.
“Many of the OEMs have wrapped their heads around this technology and the demand will grow very quickly,” Bolat says. “There are a lot of different apps that can be built on this layer of data; maybe we’ll be able to better detect sinkholes, cracks, and potholes, for instance.”
Before you get too excited about cars that will drive you to the store while you take a nap, you should know that only about 1% of the U.S. is mapped to GPR, Dr. Jeffs says. Once more vehicles are equipped with GPR, companies can use crowdsourcing to keep it current, scanning the ground underneath and comparing it to the data on file like a fingerprint database. There is a lot of scanning to be done yet.
"Mapping the road is going to be a big challenge," Dr. Jeffs says. "This is the next autonomous frontier."
Got a tip? Send the writer a note: firstname.lastname@example.org