[Paper Review] Multi-Fingered Robotic Grasping: A Primer

Stefano Carpin, Shuo Liu+2 more|arXiv (Cornell University)|Jul 22, 2016

Robot Manipulation and Learning60 references3 citations

TL;DR

This paper presents a comprehensive primer on multi-fingered robotic grasping, focusing on practical performance metrics and standardized testing frameworks. It introduces a systematic approach to evaluating robotic hands through unit, integrated, and functional tests that are independent of robot arms and perception systems, enabling objective benchmarking of grasp capabilities, force control, and sensor integration in real-world applications.

ABSTRACT

This technical report presents an introduction to different aspects of multi-fingered robot grasping. After having introduced relevant mathematical background for modeling, form and force closure are discussed. Next, we present an overview of various grasp planning algorithms with the objective of illustrating different approaches to solve this problem. Finally, we discuss grasp performance benchmarking.

Motivation & Objective

• To address the lack of standardized, objective evaluation methods for robotic hands in real-world grasping applications.
• To bridge the gap between theoretical grasping research and practical implementation by proposing performance metrics independent of robot arms and perception systems.
• To establish a framework for unit, integrated, and functional testing of robotic hands that enables fair comparison across different systems.
• To define measurable performance characteristics such as grasp force, positional accuracy, and sensor resolution for robotic hand evaluation.
• To support developers and end-users in matching robotic hand capabilities to specific application needs through a transparent, quantitative benchmarking system.

Proposed method

• Proposes a modular testing framework with three tiers: unit tests (individual components), integrated tests (hand with external forces), and functional tests (with robot arm and perception systems).
• Defines unit performance metrics including volumetric grasp capability, maximum pinch and grasp forces, finger positional accuracy, and sensor resolution (e.g., tactile force and spatial resolution).
• Introduces independent measurement systems to avoid biases from force accuracy and data latency, ensuring objective benchmarking.
• Uses standardized test objects (e.g., spheres, cylinders, cubes) to evaluate grasp configuration and force capabilities across the hand’s workspace.
• Applies kinematic and static equilibrium models (e.g., grasp matrix, wrench space, friction cone) to formalize grasp stability and force transmission.
• Emphasizes the need for generic, application-agnostic functional tests to evaluate system performance before deployment in real tasks.

Experimental results

Research questions

• RQ1 How can robotic hand performance be objectively evaluated across different systems, independent of robot arms and perception systems?
• RQ2 What standardized unit-level metrics best characterize the kinematic and force capabilities of a multi-fingered robotic hand?
• RQ3 How can sensor performance (e.g., tactile sensing) be quantitatively assessed for grasp control and feedback?
• RQ4 What integrated tests are necessary to evaluate a hand’s ability to maintain grasp under external disturbances?
• RQ5 How can functional tests be designed to validate robotic hand performance in real-world application scenarios?

Key findings

• Standardized, independent testing frameworks are essential for fair and objective benchmarking of robotic hands across different platforms.
• Unit performance tests using primitive geometries (spheres, cylinders, cubes) effectively characterize the volumetric and force capabilities of robotic hands.
• Maximum grasp and pinch forces can be measured at the bounds of volumetric capability, providing quantifiable performance indicators.
• Tactile sensor performance is best evaluated through metrics like normal and shear force resolution, directional sensitivity, and spatial resolution.
• Integrated tests that assess grasp stability under external forces reveal the robustness of a hand’s control and mechanical design.
• Functional tests involving robot arms and perception systems are critical for validating real-world applicability and should be standardized per application domain.