An innovative new system has been built by a team of researchers from MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). This system allows self-driving cars to navigate unmapped roads and drive there without the need for 3D maps.
The system is named MapLite and combines GPS data with a series of high tech sensors that observe the road conditions accurately. The system allowed the team to drive on multiple unpaved country roads in Devens, Massachusetts and to detect the road reliably more than 100 feet in advance.
Daniela Rus, Director of CSAIL at MIT said in a statement, “Cars use these maps to know where they are and what to do in the presence of new obstacles such as pedestrians and other cars. The need for dense 3D maps limits the places where autonomous cars can operate.”
MIT explains how the system is able to work without 3D maps, “MapLite uses sensors for all aspects of navigation, based on GPS data only to obtain a rough estimate of the car’s location. The system first establishes both a final destination and what researchers call a “local navigation objective, which must be in view of the automobile. Their perception sensors then generate a path to reach that point, using LIDAR to estimate the location of the edges of the road. MapLite can do this without physical markings on the road by making basic assumptions about how the road will be relatively flatter than the surrounding areas.”
During the tests on the country roads in Devens, the researchers were successful in making the car navigate along a kilometer stretch without any human intervention. The challenge now is to expand on the types of roads the system can handle. The system currently has trouble dealing with rapid elevation changes like the ones you can expect on mountain roads. The ability to follow lane markings and road signs is also being added. Once the MapLite system is perfected, it will be a huge boost for self-driving cars.