Designing Robot Exercise Partners

Co-designing how robot-supported exercise should feel for older adults.

This project began with a design question rather than a technical one: if a robot were going to support physical activity at home, what kind of relationship should it create? Across participatory workshops and a later in-home case study with Poppy Ergo Jr, the work explored how exercise support could become more social, more adaptive, and more respectful of older adults' autonomy.

HRI 2023 HRI 2024 Aging in Place Workshop Participatory Design Healthy Aging Social Robotics

Co-design paper Home deployment paper

Poppy as a robot exercise partner: a design probe shaped through participatory work with older adults and later tested in the home.

Victor Nikhil Antony*, Sue Min Cho*, Chien-Ming Huang

*equal contribution · HRI 2023 · Co-designing with Older Adults, for Older Adults

Victor Nikhil Antony, Chien-Ming Huang

HRI 2024: Workshop on HRI for Aging in Place · Designing Social Robots that Engage Older Adults in Exercise

Older adults in the co-design study, ages 65–94

Relationship models that emerged: Trainer, Companion, Augmenter

Design guidelines initially distilled from the participatory work

Days of in-home deployment with Poppy Ergo Jr

Three-stage co-design process with older adults and the broader study flow for robot exercise partners.

The project moved through interviews, workshops, and critique sessions before later testing those design ideas through an in-home deployment.

01 · Design question

The challenge was not just getting people to move. It was designing a partner they would actually want nearby.

Existing exercise tools often assume that better tracking, more reminders, or stricter routines will solve inactivity. The papers point in a different direction: older adults wanted support that understood their schedule, their limits, and the emotional texture of exercise at home. That made this project fundamentally about interaction design and relationship design.

Autonomy Scheduling

Participants wanted control over when and how they exercised, rather than daily compliance pressure imposed by the system.

Adaptation Physical limits

Pain, mobility constraints, and fall concerns made rigid exercise flows feel exclusionary instead of supportive.

Companionship Motivation

What many participants wanted most was presence: a robot that exercised with them, not one that simply monitored them.

"Having a workout group motivates me more to work out. You want to show up for the other people."

Co-design participant — on social motivation

"I'm scared of falling, and I have to be really careful with some movements. The robot should know that."

Co-design participant — on the need for physical adaptation

02 · Relationship models

Older adults designed three different kinds of robot exercise partners.

The most important output of the co-design work was not a single preferred robot. It was a small design vocabulary for how a robot could relate to a user during physical activity.

Structure

Trainer

A robot that provides routine, pacing, and clear exercise guidance for users who want an external scaffold.

Leads structured routines
Demonstrates form and counts progress
Works when guidance feels supportive, not punitive

Presence

Companion

A robot that joins the user in exercise and offers encouragement through presence rather than direction.

Exercises alongside, not above
Encourages gently without nagging
Makes movement feel shared and social

Adaptation

Augmenter

A robot that reshapes routines around physical limitations, chronic conditions, and changing mobility.

Adapts routines to the body in front of it
Prioritizes safety and accommodation
Treats variation as expected, not exceptional

03 · Artifacts

Workshops generated concrete concepts, embodiments, and interaction sketches.

The project’s design contribution lives in these artifacts as much as in the written guidelines. They make visible how participants imagined tone, form, and flow.

Prototype concepts created during co-design workshops with older adults.

Prototype concepts from workshops surfaced different expectations around posture, approachability, and social presence.

04 · Motivators and barriers

The co-design findings showed that motivation and friction are both deeply situated.

Sociality, health goals, and enjoyment motivated activity. Pain, fatigue, fall risk, and scheduling friction undermined it. Together these findings explain why the relationship model mattered so much.

Motivators for physical activity among older adults.

Motivators clustered around companionship, wellbeing, and enjoyment rather than around surveillance or optimization.

Barriers to physical activity among older adults.

Barriers were practical and embodied: pain, fatigue, fear of falling, and the difficulty of fitting routines into daily life.

Social motivation is real Companion

Participants repeatedly described wanting someone to exercise with, which gave the Companion role real grounding.

Rigid systems exclude Augmenter

Physical limitations and injury concerns made adaptation feel essential, not optional.

Respect matters Tone

The preferred interaction style was clear: encouragement without nagging, guidance without loss of control.

05 · Deployment as stress test

Poppy Ergo Jr translated the design ideas into a working home exercise system.

The later case study did not replace the co-design work. It tested it. Poppy Ergo Jr became the first probe for asking which design principles held up once the robot entered daily life.

Poppy Ergo Jr used in the home deployment study.

The home study used Poppy Ergo Jr with a touchscreen interface, hotword activation, pose estimation, and wearable sensing to structure exercise sessions in the participant's home.

Choose and begin Interface

The participant woke Poppy, selected routines through the touchscreen, and retained control over when and what to do.

Exercise with feedback System

Poppy demonstrated exercises while pose tracking and wearable data supported monitoring and coaching during the session.

See what breaks Learning

The week-long deployment revealed whether the co-designed interaction ideas were robust enough for real routines and real bodies.

06 · What held up

The deployment confirmed the value of the design direction and exposed remaining gaps.

The system worked well enough to surface a sharper insight: the missing pieces were not just technical bugs, but places where the design still reflected an engineer’s idea of exercise more than an older adult’s.

Autonomy stayed central Scheduling

Flexible routine structure mattered more than strict daily repetition, reinforcing the co-design emphasis on user control.

Variety mattered Content

A small exercise library was not enough for a week-long experience, especially given how much exercise content already exists elsewhere.

Adaptation was still incomplete Pose tracking

Rigid sensing assumptions created friction for a participant with an arm issue, validating exactly why the Augmenter role mattered.

Presence was promising Engagement

The deployment suggested that robot presence can sustain routine, but only when it respects the user’s own pacing and bodily constraints.

"I don't want to work out an hour a day — 20 minutes a day would be fine, but I'd prefer it being an hour, three days a week."

Deployment participant — on flexibility over compliance

"I have an arm issue, and unless I completed the exercise exactly the way they wanted it done, it wouldn't move forward — so I soldiered through."

Deployment participant — on why adaptation cannot be optional

07 · Takeaways

The lasting contribution is a design direction for robot exercise partners.

Across both papers, the work suggests a smaller set of durable design principles: build for autonomy, adapt to real bodies, treat social presence as a primary design material, and keep interaction accessible and legible.

Design for partnership

The robot should feel like a partner in activity, not a compliance device that enforces a routine from above.

Adapt around the body

Exercise support has to respond to pain, mobility limitations, and fluctuating capacity rather than demanding ideal form.

Support autonomy

Timing, intensity, and routine structure need to stay negotiable so the system fits into a user’s life instead of trying to reorganize it.

Keep the interaction legible

Appearance, feedback, and interface flow all shape trust. The robot has to feel approachable, understandable, and easy to use.