Typically, self-driving cars rely on a combination of cameras and other sensors to “view and confirm” their location on the road. But what if road signs are covered with heavy snow? To this end, the Massachusetts Institute of Technology has developed a system that allows cars to see snow-covered roads.
The system uses the experimental geodesic radar positioning (LGPR) technology to create a digital fingerprint underground to locate vehicles equipped with TerraVision by sending radio waves to the ground. Because the waves can penetrate three meters underground, they reflect rocks, roots, soil and pipes.
The reflected signal sits used to build a 3D map of the underground condition. Stitching together these images creates a complete 3D fingerprint that any vehicle equipped with LGPR technology can use to determine its location. In addition, the creation of complete maps, both above and below ground, increases the likelihood that self-driving cars can be positioned and navigated under any conditions.
In the closed route test, the MIT team was able to pinpoint the snow-covered road signs, with a margin of error of only about 1 inch (25 mm). Although the setting has so far been tested only at low speeds on rural roads, the researchers believe it should also be used at high speeds on highways. However, it is not intended to replace existing vision sensors in cars, but to enhance existing sensors.
The paper on the study, led by Professor Daniela Russ and PhD student Teddy Ort, will be published later this month in the journal IEEE Robotics and Automation Letters.
Source: MIT Computer Science and Intelligence Artificial Lab (CSAIL)