Grasp, Slide, Roll: Comparative Analysis of Contact Modes for Tactile-Based Shape Reconstruction

Chung Hee Kim; Shivani Kamtikar; Tye Brady; Taskin Padir; Joshua Migdal

cs.ROFeb 26, 2026

Grasp, Slide, Roll: Comparative Analysis of Contact Modes for Tactile-Based Shape Reconstruction

Authors: Chung Hee Kim, Shivani Kamtikar, Tye Brady, Taskin Padir, Joshua Migdal

Organizations: Amazon Fulfillment Technologies & Robotics, Westborough MA, USA · Robotics Institute at Carnegie Mellon University, Pittsburgh PA, USA · Siebel School of Computing and Data Science, University of Illinois at Urbana-Champaign, Champaign IL, USA

Abstract

Tactile sensing allows robots to gather detailed geometric information about objects through physical interaction, complementing vision-based approaches. However, efficiently acquiring useful tactile data remains challenging due to the time-consuming nature of physical contact and the need to strategically choose contact locations that maximize information gain while minimizing physical interactions. This paper studies how different contact modes affect object shape reconstruction using a tactile-enabled dexterous gripper. We compare three contact interaction modes: grasp-releasing, sliding induced by finger-grazing, and palm-rolling. These contact modes are combined with an information-theoretic exploration framework that guides subsequent sampling locations using a shape completion model. Our results show that the improved tactile sensing efficiency of finger-grazing and palm-rolling translates into faster convergence in shape reconstruction, requiring 34% fewer physical interactions while improving reconstruction accuracy by 55%. We validate our approach using a UR5e robot arm equipped with an Inspire-Robots Dexterous Hand, showing robust performance across primitive object geometries.

Figures & tables

Panel of Figure 1 — Fig. 1: Tactile information is promising to unlock contact-rich manipulation capabilities for robots. We investigate tactile data acquisition efficiency of various contact modes (sliding, rolling, and discrete contacts) in the context of 3D object reconstruction using tactile data only (no vision).

Fig. 2: Comparative examples of original SPVD results and our improved approach. Partial input points and shape completed output points are colored blue and red, respectively. (a) Shape completion with (right) and without (left) normal vectors (shown as black arrows) illustrates how normal guidance disambiguate shape generation direction. (b) Novel point generation filters noisy measurements (right), while the original approach preserves noisy inputs (left). (c) Rotation-invariant shape completion achieved through random rotation augmentation, demonstrated with a chair point cloud.

Panel of Figure 3 — Fig. 3: Data augmentation pipeline on a simulated tactile point cloud including contact point truncation, Gaussian noise injection, and random rotation, illustrated using contact data from a sphere. Points are colored to show correspondence before and after applying the augmentation.

Panel of Figure 4 — Fig. 4: Contact motion sequence of the tactile contact simulator. The target object is a sphere shaded in purple.

Panel of Figure 5 — Fig. 5: Process diagram of the information-theoretic tactile exploration. Measured tactile points $P_{meas}$ are colored blue (Step 1), while the shape completed point cloud $P_{pred}$ is colored red (Step 2). The coordinate frames in Step 3 represent candidate contact poses, while gray coordinate frames in Step 4 represents filtered candidates. The contact pose with the highest score is selected (Step 5) and executed (Step 6).

Panel of Figure 6 — Fig. 6: Three contact interaction modes including grasp-releasing (top), finger-grazing (middle), and palm-rolling (bottom).

Panel of Figure 7 — Fig. 7: Shape completion results on simulated tactile point clouds.

Panel of Figure 8 — Fig. 8: (a) Our system features a UR5e robot arm and the Inspire-Robots Dexterous Hands embedded with tactile sensors. (b) Tactile sensor arrays on the palmar side of the anthropomorphic end-effector is visualized in red.

Panel of Figure 9 — Fig. 9: Qualitative results of shape reconstruction on real-world tactile data. Tactile measurements are obtained through information-theoretic exploration using one of three contact interaction modes. We tested our framework on nine objects fixed relative to the robot base. † D&R : Deformed & Rotated.

Panel of Figure 10 — Fig. 10: Sequential robotic tactile exploration of a 3D-printed ball using finger-grazing motions. (Top row) Contact locations from each exploration step. (Middle row) Accumulated tactile point cloud data. (Bottom row) Progressive shape reconstruction results. See supplementary video for detailed visualizations.

Panel of Figure 11 — Fig. 11: Key metrics across different contact modes – GR: Grasp-Releasing / FG: Finger-Grazing / PR: Palm-Rolling. (a) Reconstruction Chamfer Distance: Average Chamfer distance for shape reconstruction at convergence. (b) Number of Interactions Until Convergence: Average number of interactions required to achieve shape reconstruction convergence. (c) Contact Volume per Interaction: Average volume of tactile data collected during each interaction. (d) Chamfer Distance Progression: Illustrates how reconstruction Chamfer distance decreases over successive interactions.

Shape	Variation	Dimension	Chamfer Distance			# Interactions to Converge			Contact Volume (mm 3 )
Category	Variation	(mm)	GR	FG	PR	GR	FG	PR	GR	FG	PR
Ball	Small	$r_{x}=r_{y}=r_{z}=30$	0.079	0.046	0.048	6.8	5.2	3.2	52	263	472
	Big	$r_{x}=r_{y}=r_{z}=45$	0.095	0.073	0.088	10.0	5.4	7.0	114	174	404
	D&R †	$r_{x}=r_{y}=30,r_{z}=45$	0.047	0.048	0.060	11.0	6.2	4.8	89	377	319
Box	Small	$w=h=l=60$	0.102	0.060	0.042	13.4	11.6	7.2	65	417	137
	Big	$w=h=l=90$	0.099	0.090	0.095	11.8	10.6	11.4	77	238	292

TABLE I: Quantitative Results for Real-World Tactile Shape Reconstruction

Robotic Tactile Sensing

Robotic tactile sensing equips machines to detect contact, force, slip, shape, and material properties for physical interaction.

See all 101 papers

ShapeGrasp: Simultaneous Visuo-Haptic Shape Completion and Grasping for Improved Robot Manipulation
May 4, 2026Lukas Rustler, Matej HoffmannRobotic Grasping 3D Reconstruction
Department of Cybernetics, Faculty of Electrical Engineering, Czech Technical University in Prague, Prague, Czech Republic.
Geometric Reconstruction of Extrinsic Contact Trajectories using Tactile Sensing and Proprioception for Tool Manipulation
Jun 20, 2026Seojung Min, Yoonjin Kim, Jeong-Jung Kim +1Tactile Sensing Robotic Tool Use
Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology, Daejeon 34141, Republic of Korea · Department of AI Machinery, Korea Institute of Machinery and Materials, Daejeon 34141, Republic of Korea
ZeroTouch: Tactile-Supervised Visual Contact Estimation for Contact-Rich Manipulation
Sep 18, 2026Dmitriy Kosenkov, Daniia Zinniatullina, Miguel Altamirano Cabrera +3Robotic Grasping Contact-Rich Robotic Manipulation
Intelligent Space Robotics Laboratory, Skolkovo Institute of Science and Technology, Moscow, Russian Federation.

May 4, 2026J/K move · Enter open · S save

ShapeGrasp: Simultaneous Visuo-Haptic Shape Completion and Grasping for Improved Robot Manipulation

Lukas Rustler, Matej Hoffmann

Department of Cybernetics, Faculty of Electrical Engineering, Czech Technical University in Prague, Prague, Czech Republic.

Robotic Grasping 3D Reconstruction Robotic Manipulation 3D Shape Completion Tactile Sensing for Robotic Manipulation

Open paper AskPDF arXiv

Humans grasp unfamiliar objects by combining an initial visual estimate with tactile and proprioceptive feedback during interaction. We present ShapeGrasp, a robotic implementation of this approach. The proposed method is an iterative grasp-and-complete pipeline that couples implicit surface visuo-haptic shape completion (creation of full 3D shape from partial information) with physics-based grasp planning. From a single RGB-D view, ShapeGrasp infers a complete shape (point cloud or triangular mesh), generates candidate grasps via rigid-body simulation, and executes the best feasible grasp. Each grasp attempt yields additional geometric constraints -- tactile surface contacts and space occupied by the gripper body -- which are fused to update the object shape. Failures trigger pose re-estimation and regrasping using the refined shape. We evaluate ShapeGrasp in the real world using two different robots and grippers. To the best of our knowledge, this is the first approach that updates shape representations following a real-world grasp. We achieved superior results over baselines for both grippers (grasp success rate of 84% with a three-finger gripper and 91% with a two-finger gripper), while improving the 3D shape reconstruction quality in all evaluation metrics used.

Abstract

Robotic Tactile Sensing

Explore similar work

ShapeGrasp: Simultaneous Visuo-Haptic Shape Completion and Grasping for Improved Robot Manipulation

Geometric Reconstruction of Extrinsic Contact Trajectories using Tactile Sensing and Proprioception for Tool Manipulation

ZeroTouch: Tactile-Supervised Visual Contact Estimation for Contact-Rich Manipulation