50 multimodal trajectories of a GR1 humanoid robot performing microwave opening tasks within the IsaacLab-Arena simulation environment. Each trajectory entry integrates state, vision, language, and action data to support the development of generalist robot policies.
Use Cases
- Train behavior cloning models using the action and state sequences provided in the trajectories
- Develop vision-language-action (VLA) policies by mapping language instructions to vision and action outputs
- Evaluate generalist robot policies for humanoid manipulation within the IsaacLab-Arena simulation
Strengths
- 50 high-fidelity trajectories generated in the IsaacLab-Arena environment
- Includes synchronized multimodal data streams for state, vision, language, and action
- Focuses specifically on the GR1 humanoid robot platform for manipulation tasks
- Provides full context for the specific task of opening a microwave