The IT specialist Nvidia has announced a platform for creating HD maps with autonomous vehicles.
Nvidia has an important position in the autonomous vehicle industry, not least because the company’s chips are installed in a number of vehicles. with Omniverse Nvidia also has a platform for autonomous driving and now Map comes with a platform for providing map data. This is also related to the takeover of the startup DeepMap.
Nvidia Drive Map Layer. Source: Nvidia
Precise map data is used by the vehicles to determine their position and makes autonomous driving possible in many areas. Nvidia wants to provide this data with its platform through various sources. Hundreds of thousands of kilometers of roads from China, Europe and the USA are to come together in this way – preferably motorways.
DRIVE Map is scalable, multimodal and intended to support the development of autonomous driving level 4 advance. The data comes from the AI crowd. DeepMap collected data from images and turned it into an accurate map that was constantly updated.
The accuracy of the Nvidia card platform is of enormous importance. This accuracy is also made possible by the three layers of the map. The camera layer shows lane markings, traffic lights, signs and more. The radar layer is a point cloud that aids in navigation bad weather improved. This data supplements missing map information in, for example, suburbs. The lidar layer is the most precise with an accuracy of five centimeters. Semantic data such as traffic instructions, turning lanes and zebra crossings also help. With the help of the data, the vehicles can behave in a way that can be expected from the rest of the traffic.
The Nvidia map is built using two separate map engines: the ground truth mapping engine and the crowdsourced mapping engine. The first engine collects data from survey vehicles and the crowd engine processes data from car sensors. The Ground Truth engine is based on the DeepMap Survey Map Engine technology, which has been developed over the past six years.
The data can be used by Omniverse, Nvidia’s autonomous driving development platform, to train the AI. In this way, digital twins are created from the data, which together are intended to depict a map of the world.