Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DexiTac: Soft Dexterous Tactile Gripping

Chenghua Lu, Kailuan Tang, Max Yang, Tianqi Yue, Nathan F. Lepora

TL;DR

This work introduces a reconfigurable soft pneumatic gripper with modular Rot and Dex joints and a high‑resolution DigiTac‑v1.5 tactile sensor to address grasping of objects with varied shapes. A kernel‑density based tactile image processing strategy enables real‑time grasp stability assessment, disturbance rejection, and guided dexterous manipulation, implemented via a double closed‑loop control system that separates safety pressure management from task execution. Experimental validation demonstrates distinct workspaces for different configurations, robust grasping under disturbances, and a set of dexterous tasks (pouring, rotating, mixing, twisting) across configurations. The approach offers a practical, extensible path toward soft, tactilely informed manipulation applicable to domains such as agriculture, domestic service, and sorting tasks, with potential for learning‑based perception and multi‑finger extension in future work.

Abstract

Grasping object,whether they are flat, round, or narrow and whether they have regular or irregular shapes,introduces difficulties in determining the ideal grasping posture, even for the most state-of-the-art grippers. In this article, we presented a reconfigurable pneumatic gripper with fingers that could be set in various configurations, such as hooking, supporting, closuring, and pinching. Each finger incorporates a dexterous joint, a rotating joint, and a customized plug-and-play visuotactile sensor, the DigiTac-v1.5, to control manipulation in real time. We propose a tactile kernel density manipulation strategy for simple and versatile control, including detecting grasp stability, responding to disturbances and guiding dexterous manipulations. We develop a double closed-loop control system that separately focuses on secure grasping and task management, demonstrated with tasks that highlight the capabilities above. The gripper is relatively easy to fabricate and customize, offering a promising and extensible way to combine soft dexterity and tactile sensing for diverse applications in robotic manipulation.

DexiTac: Soft Dexterous Tactile Gripping

TL;DR

This work introduces a reconfigurable soft pneumatic gripper with modular Rot and Dex joints and a high‑resolution DigiTac‑v1.5 tactile sensor to address grasping of objects with varied shapes. A kernel‑density based tactile image processing strategy enables real‑time grasp stability assessment, disturbance rejection, and guided dexterous manipulation, implemented via a double closed‑loop control system that separates safety pressure management from task execution. Experimental validation demonstrates distinct workspaces for different configurations, robust grasping under disturbances, and a set of dexterous tasks (pouring, rotating, mixing, twisting) across configurations. The approach offers a practical, extensible path toward soft, tactilely informed manipulation applicable to domains such as agriculture, domestic service, and sorting tasks, with potential for learning‑based perception and multi‑finger extension in future work.

Abstract

Grasping object,whether they are flat, round, or narrow and whether they have regular or irregular shapes,introduces difficulties in determining the ideal grasping posture, even for the most state-of-the-art grippers. In this article, we presented a reconfigurable pneumatic gripper with fingers that could be set in various configurations, such as hooking, supporting, closuring, and pinching. Each finger incorporates a dexterous joint, a rotating joint, and a customized plug-and-play visuotactile sensor, the DigiTac-v1.5, to control manipulation in real time. We propose a tactile kernel density manipulation strategy for simple and versatile control, including detecting grasp stability, responding to disturbances and guiding dexterous manipulations. We develop a double closed-loop control system that separately focuses on secure grasping and task management, demonstrated with tasks that highlight the capabilities above. The gripper is relatively easy to fabricate and customize, offering a promising and extensible way to combine soft dexterity and tactile sensing for diverse applications in robotic manipulation.
Paper Structure (16 sections, 3 equations, 12 figures, 3 tables, 1 algorithm)

This paper contains 16 sections, 3 equations, 12 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Gripper Design
  3. Design Concept
  4. Design Implementation
  5. DigiTac-v1.5 Design
  6. Gripper Control
  7. Hardware Control System
  8. Control Strategy
  9. Experimental Validation
  10. Workspace Estimation
  11. Experiment Setup
  12. Grasping Adaptability Test
  13. Specific Grasping Capabilities
  14. Disturbance Response
  15. Dexterous Manipulation
  16. ...and 1 more sections

Figures (12)

  • Figure 1: Design concept for the proposed reconfigurable gripper. (a) Fitness exercises include (a1) vertical row, (a2) pull-up, (a3) dumbbell side lift and (a4) dumbbell bicep curl. (b) The reconfigurable gripper inspired by (a1-a4) includes (b1) HookGrip, (b2) SupportGrip, (b3) ClosureGrip and (b4) PinchGrip. Jd, Jr, Ld and Lr represent the dexterous joints, rotating joints, dexterous links and rotating links, respectively.
  • Figure 2: Joints, fingers and sensor (bolts, nuts, tube excluded). (a) Rot joint, Dex joint, soft actuators and rigid connectors. Holes bordered in green are tube accesses and the rest are assembly holes. (b) Different assemblies of connectors and configurable fingers after assembly. (c) Overall view of DigiTac-v1.5 and its component design.
  • Figure 3: Reconfiguration process from Rot-Dex finger to Dex-Rot finger. (a) Unscrew the bolts. (b) Separate two joints. (c) Exchange places. (d) Enter the tube. (e) join the two joints. (f) screw the bolts.
  • Figure 4: Control System. (a) Schematic diagram of the double closed-loop control for gripper manipulation. (b) Real hardware. Two fingers' control is paralleled and achieved by multi-process, the diagram above just shows one finger's control loop as an example. More details about the multi-process control method can be found in Algorithm 1.
  • Figure 5: Image processing for DigiTac, which corresponds to steps 1-15 in Algorithm 1 (also is the details of process ① and ② in Fig.\ref{['fig_control_diagram']} )
  • ...and 7 more figures