Cupid

Cupid in action
Cupid UI
Monitoring server and robot UI

Inspriration

Around 15% of Canadians experience depression, and many university students struggle with loneliness, stress especially during intense academic periods. While campuses provide support services, students often hesitate to reach out or may simply need a small moment of connection in their daily routine.

We wanted to explore whether robotics and AI could help bridge that gap in an easy way. Instead of replacing human interaction, our idea was to create a friendly companion robot that can start simple conversations, recognize emotional cues, and offer small acts of kindness.

What it does

We built Cupid, a mobile companion robot that moves around campus and looks for students who may benefit from a friendly interaction.

The robot can:

Identify students who are sitting alone using computer vision
Autonomously approach them and ask if they would like to chat
Detect facial emotion (such as happy or sad) when the student interacts with it
Adapt its voice, tone, and conversation based on the detected emotion
Support modular add-ons, such as a candy dispenser, to create a warm and engaging experience

How we build it

Hardware

A differential-drive mobile base for autonomous movement
A smartphone acting as the camera, microphone, speaker, and display for interaction.

Software

YOLOv8 for real-time human detection
Google Gemini API for conversational intelligence
ElevenLabs API for natural and expressive voice output
A computer vision pipeline for tracking people and navigating safely
Custom user interface for interaction and control

Challenges

Integrating multiple machine learning models for human detection and emotion recognition.
Achieving smooth real-time communication between microphone input, web application, computer vision processing, LLM response, and robot hardware control.
Created an autonomous navigation stack using CV.

What we learned

Building Cupid give us experience in end-to-end integration of robotics, computer vision, and large language models in real time with practical challenges of deploying AI systems so that it be useful to humans.

What’s Next for Cupid

Improve emotion recognition accuracy and robustness in diverse lighting and environments
Expand conversational intelligence with longer-term memory and personalization
Add more interactive modules (games, wellness check-ins, campus information)
Deploy pilot testing. Don't be surprised if you see Cupid on campus one day :)