Appropriate Lengths Between Phalanges of Multijointed Fingers for Stable Grasping

Tokuji Okada, Takeo KanadeAAAI 1983

Abstract

Determining the optimal proportions of robotic finger segments is a critical design challenge for achieving versatile and stable grasping. This study introduces a geometric "wrap-up rate" ($W_r$) metric to quantify how effectively a multijointed finger encloses an object. Through extensive simulations of 28 distinct finger models grasping various 2D shapes, we demonstrate that configurations with a longer middle segment (M-type) consistently outperform other designs, including those mimicking the tapering structure of human fingers.

Introduction

As robotic end-effectors evolved from simple two-jaw grippers to sophisticated multijointed systems in the early 1980s, designers faced a new complexity: determining the ideal lengths for each phalange. Traditional stability analysis often relies on computationally intensive force-vector calculations at contact points, which may overlook the global geometric relationship between the finger and the object. This paper proposes that stability is primarily a function of "wrap-up" quality—the physical ability of a finger's kinematic chain to conform to and surround an object's geometry.

Methods

The core of this research is the development of a geometric proxy for stability:

The Wrap-up Rate ($W_r$)

The wrap-up rate is defined as the angular extent to which a finger encloses an object, calculated between the first and last contact points. A higher $W_r$ indicates a more "enveloping" grasp, which is inherently more resistant to external perturbations.

Finger Modeling and Classification

The researchers modeled a finger consisting of three main links ($L_1, L_2, L_3$), a palm link ($L_p$), and an auxiliary link ($L_0$).

Geometric model of a multijointed finger with links $L_i$ and joints $J_i$.

To explore the design space, 28 finger models were categorized into five types based on relative link proportions:

  • I-type: Distal links are longer than proximal ones.
  • D-type: Distal links are shorter than proximal ones (similar to human fingers).
  • M-type: The middle link ($L_2$) is the longest.
  • V-type: The middle link ($L_2$) is the shortest.
  • K-type: All links are of equal length.

Simulation and Datasets

The study utilized a synthetic dataset of 2D geometric primitives to evaluate the performance of each finger model.

Simulation environment featuring various 2D geometric primitives and the finger model.

The simulation environment included:

  • Circles: Radii from 25mm to 70mm.
  • Squares and Rectangles: Diagonal lengths from 50mm to 140mm.
  • Stability Criteria: A grasp was deemed stable if it maintained at least two contact points on the main links and achieved a wrap-up rate $W_r > 0.8$.

Key Results and Insights

The simulations yielded several significant findings regarding kinematic design:

  • Superiority of the M-Type: The most notable result was the consistent dominance of the M-type configuration. Fingers with a significantly longer middle link provided the best kinematic reach to "hook" around objects of various sizes.
  • Tactual Constraints: Stability requires more than just fingertip contact; the object must engage at least two finger segments to prevent it from slipping or "squirting" out of the grasp.
  • Human-like Design vs. Robotic Utility: While human fingers are D-type (tapering), the results suggest that for three-jointed robotic fingers, an M-type design offers greater versatility for common geometric shapes.

Limitations and Future Work

While providing a useful rule of thumb for finger design, this study has limitations:

  • 2D Constraint: The geometric model does not account for 3D volumes or out-of-plane forces.
  • Friction and Compliance: Idealized contact was assumed, ignoring the critical roles of surface friction and material softness in real-world grasping.
  • Kinematic Limits: The optimal proportions are sensitive to the specific angular ranges assigned to the joints.

Future research will focus on extending the wrap-up rate metric to 3D grasping and evaluating performance against more complex, irregular object geometries.

Summary updated

BibTeX
@inproceedings{okada1983appropriate,
  author = {Tokuji Okada and Takeo Kanade},
  title = {{Appropriate Lengths Between Phalanges of Multijointed Fingers for Stable Grasping}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {1983},
}
Figure