Source-linked AI summary
A Tendon-Driven Five-Fingered Hand with Distributed Tactile Perception for Dexterous Manipulation
Huayang Chen, Longhui Qin
TL;DR
Dexterous hands must support complex manipulation while addressing trade-offs among force, compliance, control, and tactile perception. This paper develops a tendon-driven five-fingered hand with hybrid construction and distributed dual-modality sensing, then evaluates it through gesture, pinching, grasping, and tactile-recording tasks. The demonstrations validate the integrated actuation-perception system and reveal complementary static and dynamic tactile responses.
Problem
Existing grippers are limited in complex manipulation, while rigid and soft hands trade off control complexity, compliance, force, and tactile sensing.
Method
The paper develops a tendon-driven, soft-rigid-hybrid five-fingered hand with strain-gauge and PVDF sensors distributed on the distal and middle phalanges of all fingers.
Results
Manipulation tasks successfully validated the hand’s dexterity and manipulation performance, while bottle-grasp signals revealed complementary static and dynamic tactile responses.
Takeaways & Limitations
The integrated hand shows potential for robotic manipulation by combining dexterous action with distributed tactile perception.
Abstract
from arXiv · showhide
To apply the techniques of embodied artificial intelligence to human-oid robots for complex manipulations, dexterous robotic hands are indispensable, which are restricted by the dexterity and tactile perception capability. In this work, we proposed a novel design of tendon-driven five-fingered hand with dis-tributed tactile perception. With a soft-rigid-hybrid structure employed, both compliance and operational force are endowed to the hand. Dual-modality tactile sensing elements are distributed on the distal and middle phalanges of all five fingers, enabling the simultaneous detection of static contact and dynamic force variations. Manipulation experiments, including counting gestures, finger-to-thumb pinching, object grasping, and bottle-grasp tactile recording, demonstrate the feasibility of the integrated actuation-perception system.
1. Introduction
Existing grippers are often limited to simple operations, while rigid and soft hands trade off force, compliance, and control complexity. The paper addresses these challenges with a tendon-driven five-fingered hand combining hybrid structure and distributed tactile sensing.
- Single-DOF grippers are restricted to simple operations on regular-shaped objects in low-complexity tasks.
- Dexterous hands are needed for in-hand manipulation, tool use, and delicate object-pose adjustment.
- Rigid hands provide manipulation capability but typically suffer from high control complexity and low compliance when grasping flexible objects.
- Soft hands offer greater flexibility but fail to provide sufficient manipulation force, making their combination with rigid structures challenging.
- The proposed hand combines tendon-driven five-fingered actuation with distributed dual-modality tactile perception across all fingers.
2. Hand Design and System Integration
The hand uses a tendon-driven, underactuated architecture spanning five fingers, palm, wrist, and forearm. Its unified control and acquisition pipeline coordinates motion commands with tactile recording, while tendon routing supports adaptive wrapping after contact.
- Overall architecture: The system comprises five fingers, a palm, a wrist module, and a forearm with integrated actuation and control hardware.The forearm placement reduces distal mass while preserving access for tendon pretensioning, wiring inspection, and hardware adjustment.
- System integration: Forearm-mounted actuator units, embedded control hardware, interface electronics, and tactile acquisition form a unified actuation-perception pipeline.Command states and tactile frames are acquired in parallel during execution.
- System integration: A state-based motion-primitive strategy groups low-level actuator targets and transition times into grasp, hold, and release phases.This organization supports repeatable examination of tendon-driven motion, contact formation, and tactile response.
- Finger mechanism: Each finger is a coupled joint chain whose posture depends on tendon displacement, routing geometry, pretension, and contact conditions.The thumb supports opposing motions, while the other fingers coordinate closure during grasping and pinch-like operations.
- Finger mechanism: Crossed-tendon, finger driving, swing, and wrist-driving paths support adaptive wrapping toward less-constrained joint directions after contact.Routing geometry and pretension also influence repeatability and contact formation.
3. Tactile Sensing and Contact Interpretation
The tactile system places strain-gauge and PVDF elements on both distal and middle phalanges of every finger. Their complementary responses distinguish static or quasi-static forces from dynamic force variations during contact.
- Sensor layout: Each finger contains four tactile sensing elements: paired strain gauges and PVDF sensors on the distal and middle phalanges.Both segments can therefore perceive static and dynamic contact forces.
- Sensor modalities: Strain gauges act as piezoresistive sensors for slow-adapting responses to static or quasi-static stimuli.Their applications include grasp and release, button pressing, object lifting, collision, and continuous contact.
- Sensor modalities: PVDF sensors act as piezoelectric, fast-adapting elements that respond to dynamic stimuli and generate voltage under changing forces.
- Sensor layout: All five full fingers receive dual-modality tactile perception rather than sensing being limited to conventional fingertip contact regions.
4. Experimental Evaluations
Manipulation demonstrations covered counting gestures, thumb-to-finger pinching, and three grasp operations, while bottle-grasp recordings examined tactile responses across contact phases. The results validate dexterous motion and complementary tactile sensing behavior.
- Dexterity demonstrations: Five counting gestures were executed by selectively flexing or extending fingers to represent numbers 1~5.Both flexion and extension movements were implemented correctly and resembled human-hand gestures.
- Dexterity demonstrations: Four thumb-to-finger pinch operations were successfully implemented through coordinated finger-driving and swing-tendon actuation.
- Dexterity demonstrations: Three grasp operations were demonstrated: five-finger bottle grasp, index–middle-finger squeezing, and thumb–index pinch-and-hold.
- Dexterity demonstrations: Successful execution across these tasks demonstrated competence in producing a large set of gestures used to constitute action primitives.
- Tactile perception: PVDF channels primarily respond to contact-state transitions, whereas SG channels vary more slowly and respond when static force changes across manipulation phases.The four-channel thumb recording spans grasp, lifting up, holding, lifting down, and release.
- Tactile perception: A time lag appears between middle- and distal-segment signals because contact with the bottle changes earlier at the middle segment.Grasping and lifting forces are superposed in signal magnitude, although force direction cannot be determined.
5. Conclusion
The tendon-driven five-fingered hand integrates distributed dual-modality tactile perception and demonstrates dexterity, manipulation performance, and bottle-grasp tactile recording across experiments.
- The hand combines tendon-driven actuation with distributed dual-modality tactile perception.Its integrated system includes mechanical design, tactile sensing, driving, and control components.
- Manipulation tasks successfully validated the hand’s dexterity and manipulation performance.The experiments demonstrated the hand’s potential for robotic manipulation.
- Bottle-grasp experiments revealed tactile perception through in-depth analysis of four-channel tactile signals.
- Future work will apply the hand to more complex applications by developing action primitives and exploiting tactile perception.The authors also propose combining embodied artificial intelligence techniques to improve operation performance and dexterity.