Exploration-assisted Bottleneck Transition Toward Robust and Data-efficient Deformable Object Manipulation

Yujiro Onishi; Ryo Takizawa; Yoshiyuki Ohmura; Yasuo Kuniyoshi

cs.ROMar 14, 2026

Exploration-assisted Bottleneck Transition Toward Robust and Data-efficient Deformable Object Manipulation

Authors: Yujiro Onishi, Ryo Takizawa, Yoshiyuki Ohmura, Yasuo Kuniyoshi

Organizations: Mechano-Informatics Department, Graduate School of Information Science and Technology, The University of Tokyo, Tokyo 113-8656, Japan.

Abstract

Imitation learning has demonstrated impressive results in robotic manipulation but fails under out-of-distribution (OOD) states. This limitation is particularly critical in Deformable Object Manipulation (DOM), where the near-infinite possible configurations render comprehensive data collection infeasible. Although several methods address OOD states, they typically require exhaustive data or highly precise perception. Such requirements are often impractical for DOM owing to its inherent complexities, including self-occlusion. To address the OOD problem in DOM, we propose a novel framework, Exploration-assisted Bottleneck Transition for Deformable Object Manipulation (ExBot), which addresses the OOD challenge through two key advantages. First, we introduce bottleneck states, standardized configurations that serve as starting points for task execution. This enables the reconceptualization of OOD challenges as the problem of transitioning diverse initial states to these bottleneck states, significantly reducing demonstration requirements. Second, to account for imperfect perception, we partition the OOD state space based on recognizability and employ dual action primitives. This approach enables ExBot to manipulate even unrecognizable states without requiring accurate perception. By concentrating demonstrations around bottleneck states and leveraging exploration to alter perceptual conditions, ExBot achieves both data efficiency and robustness to severe OOD scenarios. Real-world experiments on rope and cloth manipulation demonstrate successful task completion from diverse OOD states, including severe self-occlusions.

Figures & tables

Fig. 2: System architecture. The Object Recognition Module and Action Decision Policy handle OOD states by selecting between Preparation Action $a_{p}$ and Exploration Action $a_{e}$ , transitioning the object into a bottleneck state. Once t his transition is conducted, the Behavior Cloning (BC) Model executes the tasks from the bottleneck state.

Fig. 3: Examples of images with keypoint representations for input of the Action Decision Policy, which receives both the raw image and the image with representation. (Top row) Rope manipulation: two endpoints are extracted. (Bottom row) Cloth manipulation: represents two adjacent corners with an adjacent guideline are represented.

Panel of Figure 3 — Fig. 4: General VLM prompt structure for verification tasks.

Panel of Figure 4 — Fig. 5: Exploration Action of rope and cloth manipulation

Panel of Figure 5 — Fig. 6: Preparation Action of rope and cloth manipulation

Panel of Figure 6 — Fig. 7: Examples of bottleneck states for rope and cloth manipulation

Panel of Figure 7 — Fig. 8: OOD state examples created based on OOD protocol. Initial OOD states created by this protocol have positional diversity and complexity.

Panel of Figure 8 — Fig. 9: Examples of observations with ground-truth criteria ( $\epsilon=30$ px)

Table 9

Fig. 10: Recognizability Rate (RR) as a function of the number of Exploration Actions for rope and cloth manipulation. This metric illustrates how the integration of ADP and EA enhances object recognizability for different ORMs.

Fig. 11: Cumulative Accuracy Rate (CAR) as a function of the number of Exploration Actions for rope and cloth manipulation. This shows how ADP correctly infers recognizability throughout the recognition process.

Stage	Object	Rate	Primary Causes
Recognition	Rope	1/30	over threshold $\epsilon$
	Cloth	5/30	over threshold $\epsilon$ (1/5)
			VLM over-conservation (2/5)
			misleading inference (2/5)
Transition	Rope	6/29	Pinch grasp errors
	Cloth	2/25

TABLE IV: Summary of failure modes across different pipeline stages for rope and cloth manipulation tasks.

Explore similar work

Feb 26, 2026

A Perspective on Open Challenges in Deformable Object Manipulation

Deformable object manipulation (DOM) represents a critical challenge in robotics, with applications spanning healthcare, manufacturing, food processing, and beyond. Unlike rigid objects, deformable objects exhibit infinite dimensionality, dynamic shape changes, and complex interactions with their environment, posing significant hurdles for perception, modeling, and control. This paper reviews the state of the art in DOM, focusing on key challenges such as occlusion handling, task generalization, and scalable, real-time solutions. It highlights advancements in multimodal perception systems, including the integration of multi-camera setups, active vision, and tactile sensing, which collectively address occlusion and improve adaptability in unstructured environments. Cutting-edge developments in physically informed reinforcement learning (RL) and differentiable simulations are explored, showcasing their impact on efficiency, precision, and scalability. The review also emphasizes the potential of simulated expert demonstrations and generative neural networks to standardize task specifications and bridge the simulation-to-reality gap. Finally, future directions are proposed, including the adoption of graph neural networks for high-level decision-making and the creation of comprehensive datasets to enhance DOM's real-world applicability. By addressing these challenges, DOM research can pave the way for versatile robotic systems capable of handling diverse and dynamic tasks with deformable objects.

Ryan Paul McKennaa, John Oyekan

Department of Computer Science, University of York, Heslington, York, YO10 5DD, United Kingdom

Deformable Object Manipulation Generalization in Robotic Manipulation Deformable Object Simulation

Ask PDF

Jun 2, 2026

Instant-Fold: In-Context Imitation Learning for Deformable Object Manipulation

Deformable object manipulation (DOM) is challenging due to high-dimensional, partially observable states that evolve through long-horizon, topology-changing interactions with multiple valid manipulation modes. We introduce Instant-Fold, an in-context imitation learning framework for DOM. Given a single human demonstration, our policy infers and executes diverse manipulation modes directly from the demonstration, including variations in spatial execution and ordering, without requiring gradient updates. Our approach first learns deformation-aware visual representations via temporal contrastive pretraining, after which a flow-matching transformer policy conditioned on the demonstration predicts actions to execute the intended manipulation mode. Trained entirely in simulation, Instant-Fold generalizes across diverse folding modes and transfers zero-shot to real-world settings without additional data collection or finetuning. Videos are available at https://instant-fold.github.io.

Yilong Wang, Cheng Qian, Edward Johns

The Robot Learning Lab Imperial College London

Deformable Object Manipulation Generalization in Robotic Manipulation Imitation Learning Robot Imitation Learning

Ask PDF

May 13, 2026

SID: Sliding into Distribution for Robust Few-Demonstration Manipulation

Generalizing robotic manipulation across object poses, viewpoints, and dynamic disturbances is difficult, especially with only a few demonstrations. End-to-end visuomotor policies are expressive but data-hungry, while planning and optimization satisfy explicit constraints but do not directly capture the interaction strategies demonstrated by humans. We propose Sliding into Distribution (SID), a structured framework that learns an object-centric motion field from canonicalized demonstrations to iteratively slide the system toward the demonstrated manifold and into the reliable operating region of a lightweight egocentric execution policy, mitigating out-of-distribution (OOD) execution. The motion field provides large corrective motions when far from the demonstration manifold and naturally vanishes near convergence, enabling robust reaching under substantial pose and viewpoint shifts. Within the reached regime, an egocentric policy trained with conditioned flow matching performs task-specific manipulation, supported by kinematically consistent point-cloud reprojection augmentation that preserves action-observation consistency. Across six real-world tasks, SID achieves approximately 90% success under OOD initializations with only two demonstrations, with under a 10% drop under distractors and external disturbances. Overall, SID provides a new paradigm for few-shot manipulation: explicitly managing distribution shift via online distribution recovery.

Yicheng Ma, Wei Yu, Zhian Su +2

1Grasp Lab, Zhejiang University · 2Torch Kernel Co., Ltd.

Generalization in Robotic Manipulation Visuomotor Policy Learning Robotic Manipulation Robot Imitation Learning Robot Policy Generalization

Ask PDF