WaveSense GPR Technology Scans the Road Beneath the Vehicle

Nathan Eddy

---

22nd August 2018

A start-up founded by MIT graduates focuses on the ground beneath the vehicle with a radar system that helps avoid obstacles and stay on the path.

WaveSense announced the launch of a self-driving vehicle navigation based on ground penetrating radar (GPR), which the company claims can greatly improve navigation on all roads, even in poor weather conditions. The technology continually scans the subsurface soil layers, rocks and road bedding about 126 times per second and compares the scans to its onboard image database to determine the exact vehicle position in relation to the road.

Company CEO and co-founder Tarik Bolat said the design of the WaveSense  technology is the result of work in radar at MIT Lincoln Laboratory, first for use by the military and then for civilian vehicles. “The technology differs from traditional GPR systems in some key ways and, through trial and error, several unconventional approaches were adopted that resulted in the final product,” he commented.

Today’s self-driving vehicles rely on surface sensors that can face difficulties in inclement weather, roads with no or confusing lane markings and areas with sparse landmarks. “WaveSense measures reflections off of changes in soil type and density, roots, rocks, cavities and utility infrastructure, among a host of other things,” Bolat said. “That translates into a map that is rich in detail, stable and always available.”

By creating and using a map of the subterranean area up to 10 feet below the road, WaveSense does not rely on surface features, and so, is not subject to all of the common but challenging issues created by dependence on a surface map that must always be stable. Bolat claimed the technology has been driven over 10,000 miles and that it has demonstrated the platform is consistent and reliable, even in the toughest conditions.

The technology has demonstrated lateral accuracy in clear weather down to 4 cm (1.6 ins) and 6 cm (2.4 ins) lateral accuracy in a snowstorm in the middle of the night on snow-covered roads. Bolat noted both figures were demonstrated at highway speeds. “Because the technology is independent of and uncorrelated from the existing surface maps, if LiDAR and cameras fail it is extremely likely that WaveSense will be working and, likewise, if WaveSense fails it is extremely likely LiDAR and/or camera will be working,” he said.

The first systems were deployed in Afghanistan in 2013 in conditions where road markings were poor or nonexistent and obstruction due to poor visibility was commonplace. “We as a sector need to develop self-driving vehicles that are much safer than human-driven ones,” he said. “To do that, self-driving technology needs to use all information available to the vehicle instead of only trying to recreate how a human drives.”

— Nathan Eddy is a filmmaker and freelance journalist based in Berlin. Follow him on Twitter.