lvl5's localization system is the first and best vision-only localization package that does not need a LiDAR to achieve centimeter-level 6DoF localization accuracy.
While GPS alone can help a vehicle figure out its position in regular routing maps, the accuracy of GPS measurements alone are not enough for autonomous vehicles. Autonomous Vehicles need to robustly understand their position down to the inch. The process
of figuring out where the vehicle is positioned is called localization. lvl5's localization system takes camera feeds and CAN data as input and outputs the position of the vehicle in 6 degrees of freedom (6 DoF). In simpler terms, this
means that you can plug your vehicle cameras in to our box, and lvl5 will know exactly where the vehicle is, and where it is pointed.
To accomplish 6 DoF localization, lvl5 took an ancient navigational concept (trilaterating ego position from known constellations in the night sky) and applied similar concepts to "constellations" around us -- static man-made features like signs, curbs, and traffic lights. As the vehicle drives down the street, real-time computer vision algorithms locate these features in camera frames, comparing their position to the known positions in lvl5's Tier 1 and Tier 2 map. Since the exact positions of these features are known in lvl5's map, the accuracy of our trilateration is proportional to the number of features available in the map.