Robots Atlas>ROBOTS ATLAS
Robotics

Companion Robot

ActivePublished: 24 October 2026Published
Key innovation
Shifted a robot's primary purpose from performing a utilitarian task to sustained social-emotional interaction and relationship-building with a human.
Category
Robotics
Abstraction level
System
Operation level
SystemRobot control
Use cases
Emotional support and alleviating lonelinessElderly careDementia therapySupport for children on the autism spectrumMedication reminders and health monitoringCompanionship for children and only childrenEducation and teaching supportRobotic pets

How it works

A companion robot operates in a perceive-interpret-respond loop. Sensors (cameras, microphones, touch sensors) capture signals from the user; social-perception modules recognize face, speech, tone of voice and emotional expression. An affective-computing layer infers emotional state and selects an appropriate response. An interaction and dialogue manager sustains conversation, manages turn-taking and pursues goals (e.g. a medication reminder). Embodiment and expressive behaviours (gestures, gaze, facial expression, voice, motion) convey emotion and maintain engagement. A memory and personalization module builds a user model โ€” preferences, history and relationship โ€” enabling coherent, long-term interaction.

Problem solved

Addresses loneliness, social isolation and a care deficit โ€” especially among the elderly, children and people with disorders โ€” by providing a constant, available social-emotional presence and offloading caregivers without the burden of caring for a live animal.

Components

Social perception and multimodal sensingInput

Cameras, microphones and touch sensors that capture the user; face, speech and expression recognition.

Affective computingInterpretation

Inferring the user's emotional state and selecting the robot's appropriate emotional response.

Dialogue and interaction managerInteraction control

Natural-language conversation, turn-taking management and pursuit of interaction goals.

Embodiment and expressive behaviourOutput

A physical body plus gestures, gaze, facial expression, voice and motion that convey emotion and sustain engagement.

Memory and personalizationState

A user model โ€” preferences, history and relationship โ€” enabling coherent, long-term interaction.

Evolution

1998
Kismet (MIT) โ€” expressive robot as an experiment in affective computing
Inflection point

Cynthia Breazeal builds a robot head at MIT that recognizes and simulates emotions โ€” a foundation of social robotics.

1999
Sony Aibo โ€” first mass-market robotic pet
Inflection point

Sony launches the Aibo robotic dog (11 May 1999), popularizing the home companion-robot concept.

2001
PARO โ€” therapeutic seal publicly exhibited
Inflection point

Takanori Shibata (AIST, work from 1993) publicly exhibits the PARO therapeutic seal robot for hospital and nursing-home patients.

2009
PARO classified as a Class 2 medical device (US)
Inflection point

US regulators classify PARO as a Class 2 medical device โ€” formal recognition of a companion robot as a therapeutic tool.

2014
Pepper โ€” emotion-reading social robot

Aldebaran Robotics (SoftBank) introduces the semi-humanoid Pepper in Japan (June 2014), designed to read emotions.

2017
Jibo โ€” social home robot ships to users

Jibo (company co-founded in 2012 by Cynthia Breazeal) ships a social home companion robot for the family.

Hyperparameters (configurable axes)

Embodiment typeHigh

Robot form: zoomorphic (animal-like), humanoid or screen-based.

Target user groupHigh

Audience: elderly, children, people with disorders or general household users.

Autonomy levelMedium

Degree of the robot's autonomous action in interaction โ€” from reactive to proactive.

Interaction modalityMedium

Interaction channels: voice, touch, vision and motion.

Affective expressiveness rangeMedium

Richness of emotions the robot can recognize and express.