Self-driving cars rely on loads of advanced technology to help them see and navigate the world around them. However, researchers at MIT have built a self-driving car that doesn't rely on 3D mapping or even visible lane markings to get where it needs to go.
Instead, the team's MapLite technology enables the self-driving experience by combining basic GPS data with Lidar and Inertial Measurement Unit (IMU) sensors that collect data on angular velocity and linear acceleration. The trial was done in collaboration with the Toyota Research Institute, and the car was a Toyota Prius V.
Does it work? Researchers took the self-driving Toyota Prius V to unmarked roads in Devens, Massachusetts, where the car performed exactly as it should, reliably detecting the road more than 100 feet in advance.
The technology could be revolutionary for self-driving cars operating in rural parts of the country. Companies have to 3D map areas in vast detail to supply self-driving cars with enough information; MapLite does away with intense mapping.
“The cars use these maps to know where they are and what to do in the presence of new obstacles like pedestrians and other cars,” said Daniela Rus, director of MIT’s CSAIL. “The need for dense 3D maps limits the places where self-driving cars can operate.”
MIT researchers said the car sets a final destination goal and "sensors then generate a path to get to that point, using LIDAR to estimate the location of the road’s edges." The interim goals, called "local navigation goals," have to be within the view of the car. The technology can maneuver the car without road markings because the technology makes assumptions on road grades in the surrounding area.
The strategy is a very different one from major self-driving car companies such as Waymo or General Motors' Cruise Automation. The two have spent years mapping areas to ensure cars know all road surfaces and information prior to setting out. But, that strategy also makes it difficult to map rural areas. MapLite could bring the self-driving car revolution to the most remote areas of the world.
As you can see in the video, though, the car appears to be traveling rather slowly. It appears that, like self-driving cars themselves, MIT's solution likely has a long way to go before it is ready for commercial use as well.