Scholarlink

<p>Accurate stiffness discrimination is essential for robotic and prosthetic applications, enabling dexterous manipulation and effective interaction during activities of daily living. This study introduces a novel computational framework for perceptual stiffness classification by integrating Dynamic Time Warping (DTW)-based clustering with Analytic Hierarchy Process (AHP)-driven Multi-Criteria Decision-Making (MCDM). Fingertip force data were collected from 10 subjects during tactual grasping tasks involving 10 objects with varying stiffness levels over 6 trials. Two clustering techniques were compared: (a) K-means DTW with LB Keogh Lower Bound, and (b) K-medoids with Soft-DTW Barycentric Averaging (SDBA). The quality of the cluster is evaluated using Davies-Bouldin Index (DBI) and Silhouette Score, which revealed that K-medoids-SDBA achieved better performance (lower DBI: 0.31–0.50; higher Silhouette: 0.42–0.72). The AHP-MCDM framework synthesized multi-finger tactile clustering metrics, prioritizing compactness, separation, and consistency, to generate stiffness estimations and classify objects into predefined categories (Soft, Medium, Hard). Comparative evaluation against Support Vector Machine, Decision Tree, and Random Forest classifiers demonstrated statistically significant improvement in F1-score for the AHP-based framework, achieving a score of 97%. These findings highlight the potential of force-based perceptual stiffness classification to improve tactile sensing in prosthetics and assistive technologies. </p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Object stiffness discrimination through dynamic time warping and multi-criteria decision-making

  • 期刊论文

  • 发表日期：

    2025年7月19日

查看全文
  • Abhijit Dey,
  • Shyamanta M. Hazarika

摘要

Accurate stiffness discrimination is essential for robotic and prosthetic applications, enabling dexterous manipulation and effective interaction during activities of daily living. This study introduces a novel computational framework for perceptual stiffness classification by integrating Dynamic Time Warping (DTW)-based clustering with Analytic Hierarchy Process (AHP)-driven Multi-Criteria Decision-Making (MCDM). Fingertip force data were collected from 10 subjects during tactual grasping tasks involving 10 objects with varying stiffness levels over 6 trials. Two clustering techniques were compared: (a) K-means DTW with LB Keogh Lower Bound, and (b) K-medoids with Soft-DTW Barycentric Averaging (SDBA). The quality of the cluster is evaluated using Davies-Bouldin Index (DBI) and Silhouette Score, which revealed that K-medoids-SDBA achieved better performance (lower DBI: 0.31–0.50; higher Silhouette: 0.42–0.72). The AHP-MCDM framework synthesized multi-finger tactile clustering metrics, prioritizing compactness, separation, and consistency, to generate stiffness estimations and classify objects into predefined categories (Soft, Medium, Hard). Comparative evaluation against Support Vector Machine, Decision Tree, and Random Forest classifiers demonstrated statistically significant improvement in F1-score for the AHP-based framework, achieving a score of 97%. These findings highlight the potential of force-based perceptual stiffness classification to improve tactile sensing in prosthetics and assistive technologies.