ARISTO Hand: Sensing-Driven Distal Hyperextension for Fine-Grained Manipulation

Accepted to IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM) 2026

Abstract

Manipulating thin objects and constrained interfaces requires more than a human-like hand shape. The fingertip must align with surfaces, engage small edges, and sense subtle contact forces during interaction. The ARISTO Hand addresses this through controlled distal hyperextension and a rigid-soft fingertip sensing architecture. A sensorized fingernail provides direct force-torque feedback during edge contact, poking, and recessed interactions, while a soft capacitive tactile pad supports stable acquisition and compliant grasping. The tendon-driven finger enables distal alignment, allowing the fingertip pad to flatten against thin objects and planar surfaces. Together, these capabilities support fine-grained manipulation tasks such as SD card extraction and insertion, penny pickup, and contact-rich teleoperation.

Contributions

Controlled Distal Hyperextension

Active distal hyperextension rotates the fingertip pad into surface alignment, enabling stable contact with thin objects and planar surfaces.

Decoupled Rigid-Soft Sensing

A rigid nail-mounted force-torque sensor captures edge contact, while a compliant capacitive pad provides distributed tactile feedback.

Contact-Level Validation

Experiments validate how fingertip geometry, distal force sensing, and tactile feedback improve thin-object and constrained-interface manipulation.

SD Card Extraction and Insertion

The SD card task combines several difficult contact modes: recessed slot access, force-controlled push-to-eject interaction, thin-object retrieval, compliant acquisition, and final push-to-lock insertion. ARISTO handles these stages by switching between rigid fingernail interaction, tactile pad acquisition, and active distal hyperextension.

Insertion

Active distal hyperextension aligns the tactile pad with the card surface for stable acquisition, while the rigid nail completes the push-to-lock insertion.

Extraction

The rigid fingernail enters the recessed SD card slot and uses force-torque feedback to trigger the push-to-eject mechanism before retrieving the card.

Multi-stage SD card extraction and insertion task
The full sequence highlights the complementary roles of the rigid fingernail, tactile pad, and active distal hyperextension across extraction and insertion.

Hyperextension for Thin-Object Grasping

Thin objects are difficult to grasp because standard flexion often produces edge contact instead of stable pad contact. ARISTO uses active distal hyperextension to rotate the fingertip pad into alignment with the object surface, increasing contact area and preserving pull-out force.

Pull-out force versus block thickness with and without hyperextension
Active distal hyperextension improves pull-out force for 1–20 mm objects while retaining at least 98% of the thick-object baseline down to credit-card thickness. Standard flexion retains only 34–40% over the same thin-object range.
2.76x Mean pull-out force increase for 1-20 mm objects
>= 98% Thick-object baseline retained down to credit-card thickness
10.13 N Peak pull-out force measured without constrained motor gains

Pull-out tests were run without sensor feedback to isolate the kinematic effect of distal hyperextension. The ablation study intentionally limited forces to avoid actuator saturation during repeated trials. The 10.13 N measurement shows the higher peak pull-out force available when motor gains are not constrained.

Fingertip Sensor Design

The fingertip separates rigid edge interaction from compliant surface contact. A rigid nail-mounted force-torque sensor measures contact through the fingernail, while a soft capacitive tactile array measures distributed pressure across the finger pad.

This mechanical separation lets ARISTO distinguish nail contact, pad contact, and transitions between the two during fine manipulation.

ARISTO fingertip with force-torque sensor and NARI-Touch sensor
The distal phalanx integrates a nail-mounted force-torque sensor and a soft capacitive tactile array under the finger pad.

Sensing Fidelity

The nail-mounted force-torque sensor and proprioceptive motor effort are compared across two critical conditions: low-velocity edge contact and geometry-dependent normal force detection. Both experiments show that distal sensing is necessary for reliable fine-grained manipulation.

Edge Contact Detection

Edge contact detection comparing AIDIN FT sensor and motor effort at varying approach velocities

The nail-mounted FT sensor produces clean, repeatable contact steps across approach velocities. Motor effort is noisy and velocity-dependent. At low speeds (0.05 rad/s), stiction obscures contact events entirely.

Geometry-Invariant Force Detection

Normal force detection across joint angles comparing distal FT sensor and motor effort

The distal FT sensor tracks the target force within ±0.2 N across all joint angles. Motor effort degrades near straight configurations. At 15°, targeting 3.1 N with proprioception resulted in applying 12.3 N, a ~300% error.

Hand and Finger Design

ARISTO uses a tendon-driven finger architecture designed for controlled distal contact. Fully wrapped antagonistic tendons route around the PIP joint, idler pulley, motor pulley, and tendon tensioners, maintaining an approximately constant effective moment arm across flexion and hyperextension.

This routing enables active distal hyperextension while preserving the nominal grasping workspace, allowing the fingertip pad to align with flat surfaces without sacrificing standard flexion-based grasps.

CAD diagram of the ARISTO Hand tendon routing and distal hyperextension mechanism
The ARISTO hand combines a sensorized distal phalanx with fully wrapped antagonistic tendon routing, enabling active distal hyperextension for surface alignment and thin-object engagement.

BibTeX

@inproceedings{kim2026aristo,
  author    = {Kim, Aaron and Kang, Dong Ho and Helwig, Mark and Seo, Mingyo and Yokoyama, Kazuto and Narita, Tetsuya and Sentis, Luis},
  title     = {ARISTO Hand: Sensing-Driven Distal Hyperextension for Fine-Grained Manipulation},
  booktitle = {IEEE/ASME International Conference on Advanced Intelligent Mechatronics},
  year      = {2026},
  note      = {Accepted},
}