Action-conditional implicit visual dynamics for deformable object manipulation

Manipulating volumetric deformable objects in the real world, like plush toys and pizza dough, brings substantial challenges due to infinite shape variations, non-rigid motions, and partial observability. We introduce ACID, an action-conditional visual dynamics model for volumetric deformable objects based on structured implicit neural representations. ACID integrates two new techniques: implicit representations for action-conditional dynamics and geodesics-based contrastive learning. To represent deformable dynamics from partial RGB-D observations, we learn implicit representations of occupancy and flow-based forward dynamics. To accurately identify state change under large non-rigid deformations, we learn a correspondence embedding field through a novel geodesics-based contrastive loss. To evaluate our approach, we develop a simulation framework for manipulating complex deformable shapes in realistic scenes and a benchmark containing over 17,000 action trajectories with six types of plush toys and 78 variants. Our model achieves the best performance in geometry, correspondence, and dynamics predictions over existing approaches. The ACID dynamics models are successfully employed for goal-conditioned deformable manipulation tasks, resulting in a 30% increase in task success rate over the strongest baseline. Furthermore, we apply the simulation-trained ACID model directly to real-world objects and show success in manipulating them into target configurations.

Automated summary

This article introduces ACID, an action-conditional visual dynamics model for volumetric deformable objects. ACID integrates implicit representations and geodesics-based contrastive learning techniques to accurately represent deformable dynamics and identify state changes under non-rigid deformations. The evaluation shows that ACID outperforms existing approaches in geometry, correspondence, and dynamics predictions. The application of the ACID model in goal-conditioned deformable manipulation tasks leads to a 30% increase in task success rate over the strongest baseline. Furthermore, the simulation-trained ACID model is successfully applied to real-world objects for manipulation tasks.