VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Models

Keyphrases

• Actionable Knowledge 33%
• Closed-loop Robots 33%
• Code Writing 33%
• Composable 100%
• Dynamic Disturbance 33%
• Dynamic Model 33%
• End-effector 33%
• Form Language 33%
• Freeform 66%
• GitHub 33%
• Interaction with Environment 33%
• Language Instruction 33%
• Language Model 100%
• Large Language Models 66%
• Manipulation Tasks 66%
• Model-based Planning 33%
• Motion Primitives 33%
• Natural Language 33%
• Observation Space 33%
• Online Experience 33%
• Physical Interaction 33%
• Planning Framework 33%
• Real Robots 33%
• Rich Interaction 33%
• Robot Environment 33%
• Robot Manipulation 33%
• Robot Trajectory 66%
• Robotic Manipulation 100%
• Simulated Robot 33%
• Six Degrees of Freedom 33%
• Value Map 100%
• Vision-language Models 33%
• Zero-shot 33%

Computer Science

• Actionable Knowledge 50%
• Language Modeling 100%
• Large Language Model 100%
• Observation Space 50%