Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional Conditioning

Researchers introduced Attention-DP3, a spatially object-aware 3D diffusion policy that uses geometry-aligned attentional conditioning to address ambiguity in 3D point-cloud observations from complex, cluttered manipulation scenes. The method targets failures in standard 3D diffusion policies, which struggle to localize and exploit task-relevant geometry when target objects are partially occluded or intermingled with visually similar distractors.

3D point-cloud observations are inherently ambiguous in complex, cluttered manipulation scenes, where target objects may be partially occluded or tightly intermingled with visually similar distractors. As a result, standard 3D diffusion policies often struggle to localize and exploit task-relevant g