DeformGen: Dynamics-Based Topology Augmentation for Deformable Manipulation Policy Learning

2026-06-24

AI Summary

Existing demonstration augmentation methods struggle to generate physically plausible and topologically diverse state-trajectory pairs for deformable object manipulation, due to high-dimensional state constraints and non-equivariant trajectory transfer. This work proposes a dynamics-based augmentation framework that generates topologically consistent novel states through local physical perturbations and forward simulation, and non-rigidly transfers source manipulation trajectories to new geometries via deformation field warping, thereby jointly expanding the distribution of states and their corresponding behaviors. To the best of our knowledge, this is the first approach to achieve topology-aware data augmentation for deformable objects, overcoming the limitations of conventional rigid perturbations. Experiments demonstrate significant improvements over both raw demonstrations and rigid augmentation baselines on high-fidelity deformable manipulation benchmarks, leading to enhanced policy learning performance.

Abstract

Demonstration augmentation is proposed for cost-efficient data acquisition, but existing methods are fundamentally limited in deformable manipulation due to two challenges: (1) the state space is high-dimensional with physics-induced constraints, making valid configurations impossible to reach via low-dimensional pose perturbations; and (2) trajectory transfer is non-equivariant, as material points no longer move rigidly together under deformation. We present DeformGen, a dynamics-based augmentation framework that achieves topological diversity for deformable objects. For the state challenge, DeformGen expands the valid state distribution by applying localized physical disturbances and forward-simulating the dynamics to obtain topology-coherent, physically plausible deformable states. For the trajectory challenge, DeformGen transfers source manipulation trajectories via deformation-field warping, which lifts per-particle displacements into a continuous spatial function to adapt the end-effector trajectory consistently with the deformed geometry. In this way, our method jointly augments the state distribution and its associated manipulation behavior. Experiments on high-fidelity deformable manipulation benchmarks show that DeformGen generally improves policy learning compared with training on the original demonstrations alone and with rigid-style augmentation baselines.

Problem

• deformable manipulation
• demonstration augmentation
• state space
• trajectory transfer
• topology diversity

Innovation

• dynamics-based augmentation
• topology-coherent deformation
• deformation-field warping
• deformable manipulation
• trajectory transfer