Most of the companies test self-driving cars in big cities as they require well-labeled 3D maps to identify lanes, signs and road curbs. Those are the features that aren’t available on a country road. However, MIT might have a solution for it. Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed MapLite, a new framework that doesn’t need a 3D map to find its own way.
MapLite combines GPS using the most basic topographic maps from OpenStreetMap with IMU sensors and LIDAR, which monitor the road conditions. Researchers used a self-driving Toyota Prius and outfitted it with MapLite, LIDAR, and sensors. Without any trouble, the car was able to navigate multiple unpaved rural roads in Devens, Massachusetts just by watching them over 100 feet ahead.
MapLite isn’t perfect yet. It still has some limitations. For instance, the system doesn’t know how to cope with mountain roads and other dramatic changes in elevation. 3D maps are still quite useful if the cars need to deal with the complexity of cities, but CSAIL’s system is vital for country roads, snowy landscapes where the car needs to extemporize.
Researchers described the system in a paper, which is going to be presented at the International Conference on Robotics and Automation (ICRA) in Brisbane, Australia later this month.