Université de Sherbrooke · AMI-Lab Research

AMI-Lab Platform

A complete end-to-end research infrastructure — from smart sensors in the field to AI-powered dashboards in the cloud — built to improve quality of life through ambient intelligence.

6Research Tracks
20+Active Projects
5Platform Layers

01 / REAL-WORLD IMPACT

What the Platform Does in Real Life

The AMI-Lab platform is not just research infrastructure — it is a living system deployed in real homes, residences, and cities. Here is what it concretely enables for real people, every day.

24/7 Continuous health monitoring of older adults at home non-invasive · passive
100% Raw health data stays on-premise — never sent to the cloud federated learning · privacy
< 1 min Alert response time from anomaly detection to caregiver notification real-time · self-healing
Scalable — same platform from one apartment to an entire smart city Kubernetes · distributed

Real-World Use Cases

USE CASE 01 👵 Older Adult Living Independently at Home Senior Residence · Smart Home

An older adult lives alone. AMI-Lab sensors passively monitor movement, sleep quality, heart rate, and daily routines — without cameras or wearables. If the system detects an unusual pattern (e.g., no movement at typical morning time, abnormal vital signs), it automatically alerts the caregiver or family member in real time.

Platform flow
Z-Wave Sensors DomoSense ContextAA Anomaly Detection Caregiver Alert
USE CASE 02 🏥 Caregiver Managing Multiple Residents Senior Care Centre · Dashboard

A caregiver at a senior residence monitors 20+ residents from a single dashboard. The platform aggregates sensor data from every apartment, highlights residents with changing health patterns or alerts, and provides context-rich information — enabling faster, better-informed decisions without constant physical rounds.

Platform flow
Multi-Room Sensors Hope Server Kibana Dashboard Priority Alerts
USE CASE 03 😴 Detecting Sleep Apnea Without a Sleep Clinic Home · Non-Invasive Monitoring

A patient suspected of sleep apnea avoids costly, uncomfortable sleep clinic tests. AMI-Lab deploys a non-invasive multimodal sensor kit (oxygen saturation, temperature, humidity, heart rate) in their bedroom. The platform analyzes overnight data in real time, detects apnea events, and sends alerts to medical professionals — all from home.

Platform flow
BCG + SpO₂ Sensors DomoSense Fog Computing Clinical Alert
USE CASE 04 🚶 Helping Older Adults Navigate Their City Urban Mobility · Smart City

An older adult wants to travel across Sherbrooke by bus but struggles with icy sidewalks and steep routes. The Mobilainés platform, powered by AMI-Lab infrastructure, generates a personalized route that avoids hills, monitors real-time city conditions, and selects transit connections suited to their physical capabilities and preferences.

Platform flow
User Profile City IoT Data ContextAA Personalized Route
USE CASE 05 🧠 Supporting Patients with Alzheimer's Memory Care · Dementia Support

A person living with Alzheimer's disease experiences pain but cannot reliably self-report it. AMI-Lab deploys surface electromyography (sEMG) and physiological sensors that detect pain indicators passively. The system flags likely pain events to the care team, allowing timely intervention and improving the patient's quality of life without relying on verbal communication.

Platform flow
sEMG + CO₂ Sensors DomoSense Pain Detection AI Care Team Notification
USE CASE 06 👨‍💻 A Doctor Builds an IoT App — Without Coding Healthcare · AmI-DEU · No-Code

A geriatric physician wants to monitor frailty indicators in their patients at home but has no software development background. Using AmI-DEU's visual IDE, they define the monitoring rules in their own medical vocabulary. The framework compiles this into an autonomous agent that runs on the platform, collects the right data, and notifies the physician when thresholds are crossed.

Platform flow
Medical Expert AmI-DEU IDE IaaC Compiled Live on Platform

Who Benefits from the Platform

👵 Older Adults

Live independently and safely at home with passive, non-invasive health monitoring and smart mobility assistance.

👨‍⚕️ Caregivers & Clinicians

Monitor multiple patients remotely, receive real-time alerts, and access rich health dashboards — without manual data collection.

🔬 Researchers

Deploy IoT experiments at scale with ready-made infrastructure, federated learning support, and semantic annotation built in.

🏙️ Cities & Communities

Integrate smart city services, personalized mobility, and population-level health monitoring across entire neighbourhoods.


02 / ARCHITECTURE

Complete Platform Schema

Explore all infrastructure projects →

The AMI-Lab platform integrates five interconnected layers — from physical sensors at the edge to semantic AI engines in the cloud. Each layer is built on open, interoperable technologies enabling scalable, self-managing ambient intelligence deployments in homes, hospitals, and cities.

INTELLIGENCE LAYER — AI Frameworks, Semantic Reasoning & Learning
CLOUD LAYER — Hope Servers (Kubernetes Cluster)
DATA SERVER
Integration SvcElasticsearch · K8s
Sensing SvcMQTT · Mosquitto
OM DispatcherSandboxes
Service ManagerKubernetes master
VISU SERVER
D.Project 1–NKibana · Perm. Proxy
Main PortalKong (Nginx)
Elastic ID/SecurityX-Pack · Auth
SENTINEL SERVER ↗
Deployment SvcAnsible · GitLab CI/CD
Security SvcOpenVPN · DNS
Self-Healing SvcZabbix · Auto-restart
UPDATE SERVER
AnsibleOTA Updates
EDGE / FOG LAYER — DomoSense Gateway (Raspberry Pi)
APPLICATION LAYER — Research Projects & End-User Applications
PERCEPTION LAYER — Sensors, Devices & Physical Environment
📡 Z-WaveMotion · Door · Smoke
💙 Bluetooth/BLEMiScale · Mi Band 2
📶 LoRaOutdoor Long-Range
📱 Mobile / 5GSmartphones · Apps
🩺 WearablesFitbit · Vitals
🏠 InsteonSmart Mat · Home
🌡️ Thermal CameraNon-Invasive
🛏️ BCG Bed SensorSleep · Vitals
LEGEND
AI / Semantic / Config — click to explore
Data Processing
Visualization
Sentinel / Security — click to explore
Physical / Perception

Node Deployment Workflow

🖥️STEP 01Install NodeFlash OS image onto Raspberry Pi
🔧STEP 02Configure DomoSenseSet up EnvironmentML semantic descriptor
📡STEP 03Register on ServerAdd to Elasticsearch + configure MQTT client
🚀STEP 04Deploy NodeSentinel pushes services + self-healing activated
📈STEP 05Data & VisualizationLive data flows to Elastic → Kibana dashboards

03 / FRAMEWORK

Ambient Intelligence Development Framework

Full project page →

AmI-DEU lets domain experts — not developers — build IoT applications by describing user intentions in their own language. The framework compiles these into autonomous agents (IaaC) deployed directly on ContextAA across smart homes, hospitals, and cities.

INPUT 🧠 Domain Expert

Nurse, physician, city planner — describes intentions using a visual drag-and-drop IDE. No code required.

ENGINE ⚙️ Action Machine

Semantic compiler matches intention flows against onto-AMI and generates enriched IaaC assignments.

OUTPUT 🤖 IaaC Application

Compiled autonomous agent deploys on ContextAA — runs independently across any smart environment.

📄 Full details — visual IDE, semantic model, intention examples, and validation results — on the AmI-DEU project page.

04 / MIDDLEWARE

Context-Aware Multi-Agent Platform

Full project page →

ContextAA is the intelligent middleware at the heart of AMI-Lab — a distributed multi-agent platform that ensures service continuity as users move across smart homes, cities, and hospitals. It coordinates all IaaC agents compiled by AmI-DEU and provides the context-matching engine powering the entire platform.

🏗️ 5-Layer Architecture

Lexical → Syntactic → Reasoning → Planning → Interaction

🌍 Open Smart Space

Ambient intelligence without borders — home, city, hospital, mobile.

🤝 Agent Coordination

Deploys and coordinates all AmI-DEU IaaC mission agents in real time.

🔗 onto-AMI / UnOvi

Universal ontology enabling interoperability across all IoT sources.

📄 Full details — 5-layer architecture, open smart space model, agent coordination, and semantic reasoning — on the ContextAA project page. Also powers: Semantic Reasoning · Situation Recognition · Tyche.

05 / AI RESEARCH

F-AMAD — Federated Learning Standardization

F-AMAD is AMI-Lab's federated learning standardization group — enabling privacy-preserving machine learning across distributed IoT nodes. Raw health data stays on-premise on each DomoSense gateway; only model updates are shared with the Hope server for aggregation.

FL-AMAD 🌿 Benchmark Framework

Evaluates federated learning strategies across four dimensions — Algorithms, Metrics, Architectures, and Data Heterogeneity — in distributed IoT environments.

FL AlgorithmsBenchmarkNon-IID Data
PLATFORM INTEGRATION 🔄 Edge-to-Cloud FL

FL agents are deployed via AmI-DEU and coordinated by ContextAA. Local training runs on DomoSense nodes; global aggregation runs on Hope servers — 100% private.

On-Device TrainingPrivacy-PreservingFedAvg

07 / INFRASTRUCTURE

Platform Infrastructure

All infrastructure projects →

The Hope server cluster runs acquisition and visualization microservices in separate Kubernetes namespaces with a shared Elasticsearch backend. The Sentinel servers handle security, deployment, and self-healing across the entire node fleet.


Want to Go Deeper?

Explore individual projects, meet the team, or reach out to join AMI-Lab as a student or research collaborator.

Browse All Projects Contact Us