Skip to content
RobotWorld
← Back

Build Manipulation Policies with NVIDIA Cosmos 3 Edge

Loading video
Loading video

NVIDIA walks through preparing Cosmos 3-DROID data, post-training a 4B model, and running a manipulation policy directly on Jetson Thor with receding-horizon control, no data-center GPU in the loop.

世界模型world model具身智能NVIDIACosmosJetson Thor操作策略Twitter
Category: research
Author: @NVIDIARobotics
Date: 2026-08-19T00:00:00
Duration: 7.033s
Source: https://x.com/NVIDIARobotics/status/2090107343279116648
Reference: https://x.com/NVIDIARobotics/status/2090107343279116648

Related Videos

NVIDIA SimFoundry turns real scenes into simulation worlds
1:32

NVIDIA SimFoundry turns real scenes into simulation worlds

@yukez · 11 views · 2026-07-02
SimFoundryNVIDIA仿真
Galaxea G0.5: One Autoregressive VLA for Reasoning and Action
4:22

Galaxea G0.5: One Autoregressive VLA for Reasoning and Action

@robotsdigest · 13 views · 2026-08-13
VLAGalaxeaG0.5
τ0-VLA: Robot VLA that searches over subtask sequences at test time with a world model
3:33

τ0-VLA: Robot VLA that searches over subtask sequences at test time with a world model

@jiqizhixin · 43 views · 2026-08-07
VLA世界模型World Model
Wayve Releases GAIA-4 World Model for Robotics Simulation
0:06

Wayve Releases GAIA-4 World Model for Robotics Simulation

@alexgkendall · 6 views · 2026-08-03
世界模型WayveGAIA-4
ABot World 0.5B World Model on Consumer GPU
0:19

ABot World 0.5B World Model on Consumer GPU

@HuggingApps · 6 views · 2026-07-29
世界模型world modelABot World
Being-H0.8: Latent Tactile World Action Model
2:32

Being-H0.8: Latent Tactile World Action Model

@beingbeyond_ · 18 views · 2026-07-28
触觉感知世界模型具身智能