Belief propagation for range image mapping in autonomous machine applications

This invention helps autonomous machines, like self-driving cars, understand how things are moving around them. It takes detailed 3D scans from a LiDAR sensor, simplifies them to track how the depth of objects changes between scans, and then uses a smart algorithm called "belief propagation" to refine these depth changes. Finally, it converts this refined movement information back into a full 3D picture of how everything in the environment is moving, all performed by a processor with dedicated circuits.