US Patent Application 18311172. MAP CREATION AND LOCALIZATION FOR AUTONOMOUS DRIVING APPLICATIONS simplified abstract

MAP CREATION AND LOCALIZATION FOR AUTONOMOUS DRIVING APPLICATIONS

Organization Name

NVIDIA Corporation

Inventor(s)

Michael Kroepfl of Kirkland WA (US)

Amir Akbarzadeh of San Jose CA (US)

Ruchi Bhargava of Redmond WA (US)

Viabhav Thukral of Bellevue WA (US)

Neda Cvijetic of East palo Alto CA (US)

Vadim Cugunovs of Bellevue WA (US)

David Nister of Bellevue WA (US)

Birgit Henke of Seattle WA (US)

Ibrahim Eden of Redmond WA (US)

Youding Zhu of Sammamish WA (US)

Michael Grabner of Redmond WA (US)

Ivana Stojanovic of Berkeley CA (US)

Yu Sheng of San Diego CA (US)

Jeffrey Liu of Bellevue WA (US)

Enliang Zheng of Redmond WA (US)

Jordan Marr of Santa Clara CA (US)

Andrew Carley of Kenmore WA (US)

MAP CREATION AND LOCALIZATION FOR AUTONOMOUS DRIVING APPLICATIONS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18311172 titled 'MAP CREATION AND LOCALIZATION FOR AUTONOMOUS DRIVING APPLICATIONS

Simplified Explanation

- The patent application describes a system for generating data, creating maps using the data, and localizing to the created maps. - Mapstreams, which are streams of sensor data, perception outputs from deep neural networks, and relative trajectory data, can be generated and uploaded to the cloud. - The mapstreams are used to generate map data and a high definition (HD) map that represents data from multiple drives. - When localizing to the HD map, individual localization results are generated by comparing real-time sensor data to map data from the same sensor modality. - This process can be repeated for multiple sensor modalities, and the results can be fused together to determine a final localization result.

Original Abstract Submitted

An end-to-end system for data generation, map creation using the generated data, and localization to the created map is disclosed. Mapstreams—or streams of sensor data, perception outputs from deep neural networks (DNNs), and/or relative trajectory data—corresponding to any number of drives by any number of vehicles may be generated and uploaded to the cloud. The mapstreams may be used to generate map data—and ultimately a fused high definition (HD) map—that represents data generated over a plurality of drives. When localizing to the fused HD map, individual localization results may be generated based on comparisons of real-time data from a sensor modality to map data corresponding to the same sensor modality. This process may be repeated for any number of sensor modalities and the results may be fused together to determine a final fused localization result.