PaperlayerView exact revision on arXiv

Source-linked AI summary

ORBIT: A Modular Multi-finger Robotic Hand Platform with Rotational Finger Reconfiguration

Benito Ribadeneira, Siqi Shang, Lillian Chin

arXiv:2610.33241v1cs.RO

TL;DR

Direct comparison of multi-fingered hand morphologies under matched conditions remains uncommon, leaving morphology selection for particular manipulation tasks an open research question. ORBIT provides a modular hand whose independently rotating finger carriages reconfigure layout, and the paper compares three layouts in an initial grasping study while demonstrating orbital manipulation tasks.

  • Problem

    Direct comparisons of hand morphologies using identical hardware, control, and experimental conditions are rare, while basic design questions such as finger count remain open.

    Source: §I · p. 1

  • Method

    ORBIT uses independently rotating finger carriages on a modular hand to vary finger count, geometry, and layout, and evaluates three configurations with a three-finger instantiation.

    Source: §I · p. 2, B. Hand Instantiation · p. 3

  • Results

    In an initial 11-object grasping study, each layout performed best under different object conditions, and the hand demonstrated cube reorientation, jar-lid unscrewing, and sub-centimeter precision pinching.

    Source: §I · p. 2, §IV · p. 5, B. Object Grasping Evaluation · p. 6

  • Takeaways & Limitations

    ORBIT provides a platform for direct comparison of multi-fingered hand morphologies for specific manipulation tasks.

    Source: §I · p. 2

Abstract

from arXiv · show

Despite the popularity of multi-fingered robotic hands for dexterous manipulation, there is no systematic platform to compare different hand morphologies against one another. We present ORBIT (Orbital Reconfigurable Base-actuated Integrated Theta-motion Hand), a modular platform designed to enable principled exploration of multi-fingered hand morphology. The core contribution of ORBIT is a rotating knuckle mechanism, where independent fingers travel around a cylindrical palm via a belt drive. This enables ORBIT to perform quick recon- figuration (< 1 s) of finger layout without hand disassembly. ORBITs a robust system, with finger average angular velocity of 1.57 rad / s1 and a root finger force resistance of 50 N. Using a three fingers instantiation of ORBIT, we demonstrate an initial pilot grasping study, comparing anthropomorphic, isometric, and Barrett-style finger layouts across eleven objects of varied size, shape, and texture. We further demonstrate how ORBITs nonanthropomorphic orbital DOF enables the achievement of tasks that would be difficult to perform with fixed-knuckle designs, such as cube face reorientation, lid threading, and precision pinching of sub-centimeter objects. These results establish ORBIT as a viable platform for direct apples to apples comparison of multi-fingered hand morphologies for specific manipulation tasks

I. INTRODUCTION

Robotic hand morphologies vary, but direct comparisons under identical hardware and experimental conditions are rare. ORBIT provides a reconfigurable platform for comparing layouts and demonstrates an initial three-layout grasping study and orbital manipulation tasks.

I. INTRODUCTION

  • The three layouts differ in thumb placement and finger distribution: offset thumb, evenly distributed fingers, or centered thumb.Fig. 1 presents all three morphologies on the same ORBIT hardware.Source: Figure 1 · p. 1
  • Direct comparisons of different hand morphologies under identical hardware, control, and experimental conditions are rare, making scenario-specific morphology selection difficult.Prior work typically designs a single hand for a specific goal or studies design variables separately.Source: §I · p. 1
  • ORBIT’s independently rotating finger carriages vary finger count, geometry, and layout on the same modular hand for apples-to-apples morphology comparisons.Carriages travel around a cylindrical palm via a belt-driven V-rail system.Source: §I · p. 2
  • The paper demonstrates rapid layout reconfiguration and an initial comparison of anthropomorphic, isometric, and Barrett-style layouts using 11 objects.It also demonstrates cube reorientation, jar-lid unscrewing, and sub-centimeter precision pinching.Source: §I · p. 2

II. DESIGN

ORBIT is designed to change finger layout without rebuilding the hand and to support adding, removing, or substituting finger modules. Its rotating knuckle mechanism uses independently driven carriages traveling around a cylindrical palm.

A. Rotating Knuckle Mechanism

  • Independent finger carriages rotate around a central palm, allowing dynamic layout changes without redesigning the entire hand.The design departs from static human-like knuckles, drawing on recommendations in, and a related mechanism in.Source: §II · p. 2
  • A circular V-rail adapts a linear rail-and-carriage system so each finger can move independently around the palm.The cylindrical rail is made from two 3D-printed parts that form the V-groove.Source: A. Rotating Knuckle Mechanism · p. 2
  • Two additional opposing wheels support each carriage under the multidirectional forces and moments expected during manipulation.A belt-engaging indentation helps the carriages move around the rail without slippage.Source: A. Rotating Knuckle Mechanism · p. 3
  • Finger modules can be swapped, added, or removed on the same base architecture; adding or removing modules requires a replacement belt of appropriate length.Swapping a finger requires unplugging electrical connections and unbolting the two-screw carriage interface.Source: A. Rotating Knuckle Mechanism · p. 3, Figure 3 · p. 3

B. Hand Instantiation

The paper’s ORBIT instantiation uses three anthropomorphic fingers with cascaded four-bar linkages, a compliant palm, and a modular control system. Its finger and palm components combine mechanical features for grasping with independent control of each degree of freedom.

B. Hand Instantiation

  • The hand instantiation uses three anthropomorphic fingers, each with two four-bar linkages in series for abduction/adduction and flexion/extension.The servo-driven first loop transmits motion to the second through a gear connection.Source: B. Hand Instantiation · p. 3
  • A rigid nail extends beyond a compliant sponge fingertip, helping the finger slide under and grip flat or sub-centimeter objects.Slack, gear backlash, and sponge compression provide passive grasp compliance without force sensing.Source: B. Hand Instantiation · p. 3
  • A soft palm reproduces the thenar, hypothenar, and palmar interphalangeal prominences to support power grasping.The soft shell is 3D-printed using Silicone 40A.Source: B. Hand Instantiation · p. 3
  • Each carriage carries its own Arduino and motor driver, while independent PID position loops control the hand’s degrees of freedom.A magnetic encoder provides DOF 1 feedback; the servos use internal potentiometer feedback for DOF 2 and DOF 3.Source: B. Hand Instantiation · p. 4

III. CHARACTERIZATION

The characterization evaluates position accuracy, actuation speed, and force resistance for the rotating knuckle mechanism, as well as position accuracy and speed for the fingers.

III. CHARACTERIZATION

  • Position accuracy is one of the three metrics used to characterize the rotating knuckle mechanism.The reported results are summarized in Table I.Source: §III · p. 4
  • Actuation speed and force resistance are the other two rotating-knuckle characterization metrics.The results are summarized in Table I.Source: §III · p. 4
  • Finger characterization covers position accuracy and speed.The results are summarized in Table I.Source: §III · p. 4

A. Rotating Knuckle Mechanism

Each carriage was characterized individually with the hand fixed in a gravity-neutral orientation and the finger joints held at midpoint angles to keep inertial load consistent.

A. Rotating Knuckle Mechanism

  • Each of the three carriages was characterized individually with fingers mounted and the hand held in a fixed, gravity-neutral orientation.DOF 2 and DOF 3 were held at their midpoint angles during each trial.Source: A. Rotating Knuckle Mechanism · p. 4
  • The hand’s fixed orientation controlled for gravity-related changes during carriage characterization.The setup kept the hand gravity-neutral on the robot arm.Source: A. Rotating Knuckle Mechanism · p. 4
  • Holding DOF 2 and DOF 3 at midpoint angles maintained a consistent inertial load across trials.Source: A. Rotating Knuckle Mechanism · p. 4

1) Position accuracy:

ORBIT’s position tests showed encoder and video signals reaching commanded carriage targets within the PID dead-band after settling. The mechanism reconfigured quickly and resisted substantial finger-root forces.

1) Position accuracy:

  • After 0.6 seconds, encoder feedback and video tracking reached each commanded target within the PID dead-band.The tested motor-shaft targets were 120°, 360°, and 1080°, corresponding to carriage angles of 8.5°, 25.5°, and 76.6°; Fig. 6A plots the feedback and tracking against the target.Source: 1) Position accuracy · pp. 4–5, Figure 6 · p. 5

2) Angular velocity:

  • 1.57 rad/s was the average carriage angular velocity across the three layouts, peaking at 2.36 rad/s; the largest 76.60° transition took under 0.6 s.Any of the three evaluated layouts could be reached from another in less than one second.Source: 2) Angular velocity · p. 5

3) Force resistance.:

  • 50 N was reached before displacement in nine of ten trials; the belt slipped once at 42 N.The test increased perpendicular force until displacement exceeded the 30° dead-band or the belt slipped.Source: 3) Force resistance · p. 5

B. Hand Instantiation

The finger’s direct-drive joint showed rigid coupling between servo encoder and video measurements, while the modeled fingertip path matched tracking closely enough to plan contact configurations without per-object calibration.

1) Position accuracy:

  • DOF 2’s maximum deviation was below 2° with no measurable flex, while DOF 3’s maximum fingertip error was 5.04 mm at the extremes.The extreme-position error was attributed to accumulated joint backlash.Source: 1) Position accuracy · p. 5

2) Actuation speed:

  • DOF 2 traveled the 140° range in 0.53 ± 0.01 s, compared with 0.96 ± 0.09 s for DOF 3.DOF 3 was slower because of added linkage friction and deceleration to reach the final fingertip position.Source: 2) Actuation speed · p. 5
  • The kinematic model matched video tracking well enough to plan contact configurations without per-object calibration.Root-finger reconfiguration took under 1 s, allowing DOF 1 to serve as an active manipulation primitive.Source: 2) Actuation speed · p. 5

IV. GRASPING PERFORMANCE AND DISCUSSION

The three-finger ORBIT hand was evaluated in anthropomorphic, isometric, and Barrett-style layouts, with grip force measured using spherical and flat load-cell presentations. The anthropomorphic layout produced the highest grip force across the tested cases.

IV. GRASPING PERFORMANCE AND DISCUSSION

  • The study compared anthropomorphic, isometric, and Barrett-style layouts by fixing each finger’s root position while leaving DOF 2 and DOF 3 free for grasping.The root angles were (20°, 60°, 270°), (0°, 120°, 240°), and (345°, 75°, 210°), respectively.Source: §IV · p. 5

A. Grasping Force

  • Grip force was measured with a 25 mm spherical load cell centered at the fingertips and a 20 mm × 40 mm flat load cell pressed by as many fingers as possible.The isometric flat-cell test used two opposing fingers because a three-finger grasp caught the cell edges and gave falsely weaker results.Source: A. Grasping Force · p. 5, Figure 7 · p. 6
  • The anthropomorphic configuration presented the highest grip force across the three layouts because its root-finger separation directed force more vertically toward the gauge center.The comparison used spherical and flat load-cell presentations after a 10 s settling window; the flat presentation reported higher forces because all finger forces were perpendicular to the gauge.Source: A. Grasping Force · pp. 5–6, Table II · p. 6

B. Object Grasping Evaluation

Across 11 objects, the three finger layouts showed different strengths: anthropomorphic for unevenly weighted objects, isometric for large symmetric objects, and Barrett for centered tools. All layouts grasped several everyday objects, while a heavy box and a fine-pinch screw challenged them.

B. Object Grasping Evaluation

  • For each of 11 objects, the team lowered the arm, closed the fingers to an object-specific position, lifted 100 mm, and scored stable three-second holds as successes.Source: B. Object Grasping Evaluation · p. 6
  • Anthropomorphic layout performed best on unevenly weighted objects, isometric on large symmetric objects, and Barrett on centered tools.The anthropomorphic thumb resists overturning moments; isometric contacts balance grip force; Barrett’s centered thumb levels tool pick-ups.Source: B. Object Grasping Evaluation · p. 6
  • All three configurations successfully grasped the Pringles bottle, hat, earbud, glass, and drill.The study tested 11 objects, including five 2 cm cubes and a guiro.Source: B. Object Grasping Evaluation · p. 6, Figure 8 · p. 6
  • The 430 g cracker box exceeded fingertip friction and servo grasp force, while the M3x5 mm screw required a fine pinch unavailable with fixed finger roots.Source: B. Object Grasping Evaluation · p. 6

C. Qualitative Manipulation Demonstrations

The demonstrations use ORBIT’s orbital finger motion as an active manipulation primitive while fingers remain in contact. They show dynamic screw pinching, cube reorientation, and jar-lid unscrewing, alongside contributions from modular fingertip and palm features.

C. Qualitative Manipulation Demonstrations

V. CONCLUSIONS AND FUTURE WORK

ORBIT is a modular hand platform for exploring morphology through dynamic finger reconfiguration. The authors use it to compare three established layouts and demonstrate manipulation behaviors difficult with fixed-knuckle designs; future work targets force control, sensing, and simulation.

V. CONCLUSIONS AND FUTURE WORK

  • ORBIT combines a rotating-knuckle mechanism with a two-DOF linkage-based finger design, and its dynamic reconfigurability supports direct comparison of three established hand morphologies.Source: §V · p. 7
  • Dynamic finger positioning during manipulation achieved behaviors described as normally difficult with purely anthropomorphic motions.Source: §V · p. 7
  • Future work plans force-controlled grasp commands with tactile sensing, software simulation of hand-design variables, and physical validation of sim-to-real transfer.Source: §V · p. 7
Loading 2610.33241v1…