MIT Researchers Working on Way to Make Autonomous Cars Better in Inclement Weather

Home > News > Content

【Summary】LiDAR, cameras, and sensors can be rendered useless by poor weather, but researchers at MIT believe that they have a new system that can fix the issue.

Original Vineeth Joel Patel    Apr 08, 2020 7:00 AM PT
MIT Researchers Working on Way to Make Autonomous Cars Better in Inclement Weather
Photo: WaveSense

Autonomous cars aren't on the road yet, but that hasn't stopped automakers from preparing for the future. Manufacturers are slowly rolling out futuristic technology to modern cars to see how everything behaves. So, you'll be able to find LiDAR, sensors, and multiple cameras on cars running around the road. For companies and automakers testing autonomous cars, inclement weather is a major inconvenience. Snow, dirty snow, and rain can make things difficult for all of the hardware to get a clear picture of the road. Researchers at MIT believe they have a solution.

Looking Through The Ground

A team from MIT's Computer Science and Artificial Intelligence Lab (CSAIL) has come out with a system that allows autonomous vehicles to see through the ground. It's built on technology that's called "ground-penetrating radar" (GPR) to send electromagnetic pulses underneath the road the vehicle is driving on. This is done to measure a specific area's mixture of roots, rocks, and soil.

This seems like a radical thing to do to just see the road. But MIT claims that it's done to a unique fingerprint of the road that vehicles can use later to familiarize themselves with that specific plot of land. This helps the autonomous get around needing cameras or lasers as a way to navigate.

"But LGPR can quantify the specific elements there and compare that to the map it's already created, so that it knows exactly where it is, without needing cameras or lasers," said CSAIL PhD student Teddy Ort.

During the researchers' tests, they found that the system's margin of error in snowy conditions was only an inch off compared to when it was operating in clear weather. Rainy conditions proved to be even more troublesome, as the system was off by roughly 5.5 inches. With the system installed on an autonomous vehicle, MIT researchers claimed that they never had to unexpectedly grab the steering wheel in testing. MIT only tested the system on closed country roads at low speeds, though.

How LGPR Compares To LiDAR

"Our work demonstrates that this approach is actually a practical way to help self-driving cars navigate poor weather without actually having to be able to ‘see' in the traditional sense using laser scanners or cameras," said MIT professor Daniela Rus, senior author on the new paper.

As MIT claims, this is the first time that companies looking into autonomous vehicles have utilized ground-penetrating radar as a way to help self-driving cars see. Before, the technology was predominantly used in lunar exploration, landmine detection, and construction planning.

Beyond giving autonomous vehicles the ability to immediately see what's in the ground, there's a long-time benefit to the system, too. The researchers believe that areas that are mapped out with LGPR tend to hold up better over time than maps that have been created with LiDAR. Data wise, maps made with LGPR also take up approximately 20 percent less space than a traditional 2D sensor map.

Unfortunately, the innovative system isn't ready for the market. The hardware needed to give autonomous cars to see the road are massive, roughly six feet wide. Then, there's the issue of allow LGPR datasets to be blended together to be able to handle multi-lane roads. Still, it's an innovative way around an issue that everyone will face.

Prev                  Next
Writer's other posts
    Related Content