Learning to Singulate Layers of Cloth using Tactile Feedback

Tirumala, Sashank; Weng, Thomas; Seita, Daniel; Kroemer, Oliver; Temel, Zeynep; Held, David

doi:10.1109/iros47612.2022.9981341

Cited by 18 publications

(10 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While some prior work has studied singulating a single sheet or fabric layer from a stack, most use tactile sensing or specialized end effectors. Tirumala et al [3] use a ReSkin sensor [2] to singulate layers of cloth from tactile feedback. Manabe et al [21] design a rolling hand mechanism to separate a single sheet from a pile of fabrics.…”

Section: A Deformable Objects and Single-layer Graspingmentioning

confidence: 99%

“…Given the model classification, SLIP adjusts the gripper height and retries the grasp if it does not successfully grasp a single layer. We choose to use a fixed height adjustment each time which is similar to the strategy in [3], with height deltas ∆h − and ∆h + . One could also choose to let the adjustment height decay over time or use the bisection method, but we empirically find the numbers of trials these approaches take are similar while a fixed height adjustment is more robust to a long sequence of iterations than a decaying step size, and more robust to minor bag state changes between grasps and model classification errors than bisection, since once bisection enters into a wrong interval, it may not succeed.…”

Section: Slip: Singulating Layers Using Interactivementioning

confidence: 99%

“…Such tasks are very challenging for robots. Enabling touch sensing may require equipping the robot end effector with compliant grippers or special tactile sensors such as the mini-Delta gripper [1] and the ReSkin sensor [2] used in Tirumala et al [3] and GelSight [4] used in Sunil et al [5] for cloth manipulation.…”

Section: Introductionmentioning

confidence: 99%

“…An alternative approach is to open the bag and insert objects horizontally, but this requires grasping only the top layer to create the bag opening. This is nontrivial because: (1) The depth values from a typical RGBD camera are often severely inaccurate on bags due to reflection and transparency; (2) even if the depth of the top layer is known, moving the gripper to that height to perform the grasp will push the surface downward and cause a missed grasp due to lack of friction on the surface; and (3) grasping from the side is not always kinematically feasible since the two layers can be stuck together with no space in between. As we demonstrate in experiments, a 1 mm change in the gripper height can lead to the difference between a missed (0-layer) grasp, a 1-layer grasp, and a 2-layer grasp of plastic bags, and the heights of successful 1-layer grasps are different each time depending on the shape of the bag (wrinkles and flatness of both the top and bottom layers).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Efficiently Learning Single-Arm Fling Motions to Smooth Garments

Chen

Huang

Novoseller

et al. 2023

Springer Proceedings in Advanced Robotics

View full text Add to dashboard Cite

Many fabric handling and 2D deformable material tasks in homes and industry require singulating layers of material such as opening a bag or arranging garments for sewing. In contrast to methods requiring specialized sensing or end effectors, we use only visual observations with ordinary parallel jaw grippers. We propose SLIP: Singulating Layers using Interactive Perception, and apply SLIP to the task of autonomous bagging. We develop SLIP-Bagging, a bagging algorithm that manipulates a plastic or fabric bag from an unstructured state, and uses SLIP to grasp the top layer of the bag to open it for object insertion. In physical experiments, a YuMi robot achieves a success rate of 67% to 81% across bags of a variety of materials, shapes, and sizes, significantly improving in success rate and generality over prior work. Experiments also suggest that SLIP can be applied to tasks such as singulating layers of folded cloth and garments. Supplementary material is available at https://sites.google.com/view/slip-bagging/.

show abstract

Section: A Deformable Objects and Single-layer Graspingmentioning

confidence: 99%

Section: Slip: Singulating Layers Using Interactivementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Efficiently Learning Single-Arm Fling Motions to Smooth Garments

Chen

Huang

Novoseller

et al. 2023

Springer Proceedings in Advanced Robotics

View full text Add to dashboard Cite

show abstract

“…Further, although Foldsformer can successfully perform tasks on clothes of different configurations with a single general demonstration, the demonstration sub-goals still need to be generated by expert demonstrations. In future work, we will try to make a fusion of visual information and tactile signals [31] to improve the grasp quality and reliability. We will also explore generating the sub-goals automatically.…”

Section: Limitations and Future Workmentioning

confidence: 99%

Foldsformer: Learning Sequential Multi-Step Cloth Manipulation With Space-Time Attention

Mo,

Xia,

Wang

et al. 2023

Preprint

View full text Add to dashboard Cite

Sequential multi-step cloth manipulation is a challenging problem in robotic manipulation, requiring a robot to perceive the cloth state and plan a sequence of chained actions leading to the desired state. Most previous works address this problem in a goal-conditioned way, and goal observation must be given for each specific task and cloth configuration, which is not practical and efficient. Thus, we present a novel multistep cloth manipulation planning framework named Foldformer. Foldformer can complete similar tasks with only a general demonstration and utilize a space-time attention mechanism to capture the instruction information behind this demonstration. We experimentally evaluate Foldsformer on four representative sequential multi-step manipulation tasks and show that Foldsformer significantly outperforms state-of-the-art approaches in simulation. Foldformer can complete multi-step cloth manipulation tasks even when configurations of the cloth (e.g., size and pose) vary from configurations in the general demonstrations. Furthermore, our approach can be transferred from simulation to the real world without additional training or domain randomization. Despite training on rectangular clothes, we also show that our approach can generalize to unseen cloth shapes (T-shirts and shorts). Videos and source code are available at: https://sites.google.com/view/foldsformer.

show abstract