1 of 18

Lessons Learned from Designing and Evaluating a Robot-assisted Feeding System for Out-of-lab Use

Author List

VENUE

Presenters: Name Lastname (summary), Name Lastname (moderator)

CSE 590 K: Robotics + LLMs Reading Group, Autumn 2023

Sept 27, 2023

2 of 18

Motivation

  • Millions of people cannot independently due to disability.
  • Caregiver-assisted meals may cause:
    • Self-consciousness
    • Pressure
    • Feeling like a burden

​

  • Robot-Assisted Feeding
    • Offers autonomy and dignity to people with motor impairments.
  • Most existing research:
    • Focuses on specific subcomponents
    • Conducted in controlled lab settings only
  • Goal: Develop an end-to-end, out-of-lab robot-assisted feeding system that enables people with motor impairments to independently eat full meals

3 of 18

Approach/Method

  • Co-Design: Built with input from two community researchers (CRs) with quadriplegia using Community-Based Participatory Research (CBPR) principles
  • System Design Features
    • Customizable Web App
    • User-in-the-Loop Bite Selection
    • Portable Hardware
    • Flexible Control Architecture
    • Multiple Levels of Autonomy
  • Key Design Principles : Portability, Safety, Reliability, Customizability, User Control

4 of 18

System

5 of 18

Study 1: Multi-User, On-Campus Study

  • Quantitative
  • How does the system perform across different users in out-of-lab settings?
  • 5 users and 1 CR(2)
  • Locations : Cafeteria, office, conference room

​

6 of 18

Study 2: Single-User, In-Home Deployment

  • Qualitatively
  • How does the system perform across the diverse contexts that arise when eating in the home?
  • 1 CR(2) with quadriplegia
  • Duration: 5 days, 10 meals
  • Locations: Bed & wheelchair setups at home

7 of 18

Findings

  1. Spatial contexts are numerous, customizability lets users adapt to them.
  2. Off-nominals will arise, variable autonomy lets users overcome them.
  3. Assistive robots’ benefits depend on context.

8 of 18

Finding 1: Spatial contexts require customization

  • Home environment more complex than lab
  • Users must frequently adjust robot parameters and poses

​

  • Tinkering is vital
  • Systems should provide
    • Intuitive parameter tuning
    • Clear feedback on the effects of changes

9 of 18

Finding 2: Off-nominal events are inevitable

  • Failures occurred due to lighting, planning, hardware shifts, etc.
  • Users recovered through teleoperation or mixed manual/autonomous control

​

  • Autonomy must be adjustable
    • Users can still benefit from flexible control options
    • Control can be worth the cognitive effort if empowerment is preserved

10 of 18

Finding 3: Effectiveness varies by context

  • Success depended on activity, social, temporal, and food context
  • Most effective in relaxed settings (e.g., dinner)
  • Less effective in task-heavy or time-sensitive contexts

​

  • Assistive robots provide value only in certain contexts
  • Contex-limited benefits are still meaningful and empowering

​

11 of 18

Limitations & Future Work

  • Bite acquisition
  • Bite transfer
  • Customizability
  • User control & debugging
  • User comfort
  • Cost & commercial viability
  • Caregiver integration

​

  • Long-term vision

12 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname

13 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname

14 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname

15 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname

16 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname

17 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname

18 of 18

Questions

  • Add your questions here as a bullet list

Name Lastname