Care Copilot

I started this project as an entry for the M5Stack Global Innovation Contest 2025, driven by a personal experience of caring for elderly family members. I was troubled by how existing monitoring systems often feel invasive and strip away dignity—cameras watching every move, constant alarms causing "alert fatigue, " and restrictions that prevent independent activities.

The goal was to create something different: a monitoring system that respects dignity while ensuring safety. That's how Care Copilot was born—not just a "monitor, " but a "copilot" that supports independent living.

The core innovation is Intelligent Alert Filtering. Instead of bombarding caregivers with every minor event, Care Copilot uses multi-sensor fusion and AI to determine what truly needs attention. The system aims to reduce meaningless notifications by 80%, allowing caregivers to focus on what matters.

Privacy-first design was non-negotiable. Rather than using cameras, Care Copilot relies on ToF distance sensors, environmental monitoring, and motion detection. This creates a comprehensive safety net without the feeling of being watched. The ToF4M sensor detects bed exits and falls, the ENV.4 sensor monitors room conditions, and the built-in IMU detects sudden movements indicating falls.

Currently implemented features include:

• Fall detection: Using IMU to detect >2.5G acceleration changes
• Bed exit detection: ToF sensor monitoring >2m distance changes
• Environmental monitoring: Temperature and humidity alerts
• Real-time notifications: LINE Messaging API integration for instant alerts
• GPS tracking: Outdoor location monitoring for safety

Real-world testing has shown promising initial results:

• Fall detection accuracy: Successfully detecting sudden movements
• Response time: < 3 seconds from event to notification
• MQTT reliability: 99% message delivery success
• Setup time: Under 30 minutes for complete installation

Current development status (60% complete): ✅ Completed:

• All sensor integration (IMU, ToF, ENV.4, GPS)
• MQTT communication to cloud
• LINE bot for bidirectional messaging
• Basic alert system

⏳ In development:

• Extended battery optimization (targeting 8+ hours continuous operation)
• M5StickC Plus2 integration for wearable outdoor monitoring
• PostgreSQL database for data persistence
• Web dashboard for real-time monitoring
• Sleep pattern analysis
• Daily QOL reports

One of the exciting future features is the "Independent Walk Mode" using M5StickC Plus2. This wearable device will enable elderly residents to safely go outside alone, with GPS tracking and automatic alerts only when necessary. This promotes both physical health and mental well-being while maintaining safety.

The system is designed to be affordable (under $200) and easy to deploy in small-to-medium care facilities that cannot afford expensive commercial solutions. All code is open-source, encouraging community improvements and customization.

Whether deployed in group homes, assisted living facilities, or private residences, Care Copilot aims to transform elderly care from "surveillance" to "support, " ensuring safety without sacrificing dignity.