“Holes” in maps have created many problems for self-driving cars. Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) created RoadTracer, an automated means of building road maps with more accuracy than previous approaches.
By using data from aerial images road maps can be more accurate and cost-effective. MIT professor Mohammad Alizadeh said that tech companies such as Google will be key candidates for RoadTracer.
RoadTracer can help navigate and display roads where maps are out-of-date. Looking at aerial images of New York City, RoadTracer correctly mapped 44 percent of road junctions. This is more than twice effective than current approaches, which can map about 19 percent based on image segmentation.
For the rest of the story, click here.