AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


Exploring Shared Control to Improve Self-Sufficiency of Tetraplegics: Assistive Robotics using Multimodal Intent Prediction

Authors

;

Term

4. term

Publication year

2019

Pages

121

Abstract

Personer med tetraplegi (lammelse i alle fire lemmer) har ofte begrænsede muligheder for at klare aktiviteter i dagligdagen (ADL) uden hjælp. Dette projekt undersøger et delt kontrolsystem, som blander brugerens input med automatisk assistance ved hjælp af computersyn og flere typer signaler for at forudsige brugerens hensigt. Efter en gennemgang af tidligere arbejde designede vi en løsning, der støtter simple handlinger, som er nødvendige i dagligdagen. Systemet bruger galvanisk hudrespons (måling af hudens ledningsevne) og en ny metode, der forudsiger hensigt ud fra tidligere brugerinput. Vi evaluerede systemet med 24 raske deltagere og indsamlede både subjektiv feedback og objektive præstationsdata. Resultaterne tyder på, at aggressiv arbitrering—når systemet i høj grad vægter sine egne beslutninger—kan forringe præstationen på nogle mål. Der er dog en afvejning mellem hjælp og brugerens kontrol, som kræver mere arbejde og længere, mere omfattende evalueringer for at blive afklaret.

People with tetraplegia (paralysis affecting all four limbs) often have limited ability to perform activities of daily living (ADL) without help. This project explores a shared control system that blends user input with automated assistance, using computer vision and multiple signals to predict the user’s intent. After reviewing prior work, we designed a solution to support simple actions needed in daily life. The system uses galvanic skin response (a measure of skin conductance) and a new intent prediction approach based on previous user input. We evaluated the system with 24 able-bodied participants, collecting both subjective feedback and objective performance data. The results suggest that aggressive arbitration—when the system strongly prioritizes its own decisions—can hinder performance on some measures. However, there is a trade-off between assistance and user control that requires more work and longer, more comprehensive evaluations to define.

[This abstract was generated with the help of AI]