Design and Development of a Robotic Hand for Dexterous Manipulation
Authors
Jensen, Peter Plass ; Jakobsen, Emil Faldt ; Lauridsen, Anders Bloch
Term
4. term
Education
Publication year
2026
Abstract
This project builds and tests a five-finger robotic hand designed to grasp and manipulate objects in a human-like way. The hand is underactuated (it has fewer motors than joints) and tendon-driven (cables act like tendons). Its mechanics are deliberately compliant, and the design uses morphological computation—letting the hand’s shape and materials handle part of the adaptation—so sensing and control can stay relatively simple. The system has seven actuated degrees of freedom, optimized tendon routing, force feedback, and a software control architecture with a graphical interface for composing gestures and motions. We evaluated the hand with grasping, force application, and manipulation tests. It produced stable power and intermediate grasps across many object types while adapting to their shapes. A key limitation was precision: precision grasps were less reliable because fingertip trajectories could not be controlled finely. Overall, the results show that a tendon-driven, underactuated design can deliver useful, human-like grasping without highly complex sensing and control.
Dette projekt bygger og afprøver en fem-fingret robothånd, der kan gribe og manipulere objekter på en menneskelignende måde. Hånden er underaktueret (den har færre motorer end led) og senedrevet (kabler fungerer som sener). Mekanikken er bevidst eftergivelig, og designet udnytter morfologisk computation—håndens form og materialer hjælper med at tilpasse grebet—så styring og sensorer kan holdes relativt simple. Systemet har syv drevne frihedsgrader, optimerede seneforløb, kraftfeedback samt et styresystem med grafisk brugerflade til at sammensætte håndstillinger og bevægelser. Vi evaluerede hånden gennem gribe-, kraft- og manipulationstests. Den opnåede stabile kraft- og mellemgreb på mange objekttyper og tilpassede sig deres form. En vigtig begrænsning var præcision: præcisionsgreb var mindre pålidelige, fordi fingerspidsernes baner var svære at styre fint. Samlet set viser resultaterne, at et senedrevet, underaktueret design kan levere nyttige, menneskelignende greb uden meget komplekse sensorer og styresystemer.
[This abstract has been rewritten with the help of AI based on the project's original abstract]