US 12,188,779 B2
Keyframe-based compression for world model representation in autonomous systems and applications
Akash Chandra Shekar, Redmond, WA (US); Matthew Ashman, Redmond, WA (US); and Vaibhav Thukral, Bellevue, WA (US)
Assigned to NVIDIA CORPORATION, Santa Clara, CA (US)
Filed by NVIDIA Corporation, Santa Clara, CA (US)
Filed on Mar. 10, 2022, as Appl. No. 17/654,382.
Prior Publication US 2023/0288223 A1, Sep. 14, 2023
Int. Cl. G01C 21/30 (2006.01); G01C 21/00 (2006.01); G05D 1/00 (2006.01); G06V 20/56 (2022.01)
CPC G01C 21/3819 (2020.08) [G01C 21/30 (2013.01); G05D 1/0219 (2013.01); G06V 20/588 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
determining, at a first frequency and based at least on map data, a plurality of keyframes corresponding to a world model that is representative of one or more areas around a vehicle; and
generating, at a second frequency greater than a first frequency, a plurality of world model frames corresponding to the world model based at least on a particular keyframe of the plurality of keyframes and a plurality of localization results determined with respect to the vehicle, the generating of individual world model frames of the plurality of world model frames including spatially transforming the particular keyframe to a coordinate frame of reference of the vehicle based at least on an individual localization result of the plurality of localization results.