Learning Object-Centric Spatial Reasoning for Sequential Manipulation in Cluttered Environments

Researchers developed Unveiler, a robotic manipulation framework that uses object-centric spatial reasoning to retrieve items from cluttered environments. The system achieves up to 97.6% success in simulation by separating high-level spatial reasoning from low-level action execution, and demonstrates zero-shot transfer to real-world scenarios.

• →Unveiler framework separates spatial reasoning from action execution for more efficient robotic manipulation in cluttered environments.
• →The system achieves 97.6% success in partially occluded and 90.0% in fully occluded scenarios during simulation testing.
• →The approach is more computationally efficient than large-scale end-to-end models while maintaining superior performance.
• →Spatial reasoning components transfer zero-shot to real-world scenarios without retraining learned components.
• →The framework uses a two-stage training approach combining imitation learning and PPO fine-tuning for optimal performance.