Software Engineering · AMI-Lab

AmI-DEU

Ambient Intelligence — Domain Expert User Framework

A semantic framework that empowers domain experts — not developers — to build intelligent IoT applications by describing user intentions in their own language, and automatically compiling them into autonomous agents deployed on smart environments.

Intention as a Context (IaaC) Visual IDE · No-Code Semantic Processing ContextAA Integration Smart Home · Smart City
Explore on Platform ↗
Project Snapshot
TrackSoftware Engineering
Core outputIaaC Applications
Target userDomain Expert
PlatformContextAA
IDE modelFlow metaphor
Ontologyonto-AMI / UnOvi
Validated inSmart Home · Smart City
01 / The Problem We Solve

Building IoT apps today requires software engineers.
AmI-DEU removes that dependency.

A geriatric physician, an occupational therapist, or a city mobility planner all have rich, domain-specific knowledge about what an IoT application should do — but none have the programming skills to build it. AmI-DEU bridges that gap by letting domain experts express their knowledge as activity intentions, which the framework automatically compiles into deployable smart environment applications.

"AmI-DEU empowers domain experts to describe applications in diverse domains based on the user intention — an anticipated outcome of the end-user goal represented as a self-described Context."

— AmI-DEU Framework · AMI-Lab Research · Université de Sherbrooke
🧠 Domain Knowledge In, Smart App Out

Domain experts describe what should happen in their own vocabulary. The framework handles the how.

♻️ Portable Across Environments

A single intention definition runs in a smart home, a hospital, and a smart city — without modification.

🔒 Preserves Original Intent

IaaC applications carry enriched, compiled knowledge. The original meaning is never lost during compilation or deployment.

Autonomous Once Deployed

Deployed agents run autonomously on ContextAA. They assess conditions, act on the environment, and adapt — no human in the loop.


02 / How It Works

From Intention to Deployed Agent

AmI-DEU follows a three-phase lifecycle: domain experts describe their domain, express activity intentions as flows, and the action machine compiles IaaC applications for autonomous deployment.

PHASE 01 👤 Domain Expert Geriatrician, nurse, city planner, occupational therapist…
PHASE 02 🗂️ Define Domain Concepts, attributes and relations (e.g., House, Person, Rooms)
PHASE 03 🎨 Build Intention Flow Drag & drop Entity → IF condition → Action blocks
PHASE 04 ⚙️ Action Machine Semantic compiler · onto-AMI matching · IaaC generation
PHASE 05 🤖 IaaC Application Compiled agent with enriched conditions, actions and knowledge
PHASE 06 🌐 ContextAA Deploy Runs autonomously across home, city and clinical environments

The AmI-DEU Visual IDE

A drag-and-drop environment built on the flow metaphor — no code, no terminal, no configuration files. Two separate workspaces reduce cognitive load: one for defining the activity domain, one for building application flows.

AmI-DEU — Application Designer
DOMAIN Concept → MQTT API INTENTION Entity New IF • Act •
MQTT · DomoSense Person — Victor IF · Heart Rate > 110 Act · Notify Caregiver Act · Reduce Activity
↑ Example: A health monitoring intention designed in the AmI-DEU visual IDE — no code written

03 / Semantic Model

Three Core Semantic Elements

AmI-DEU's semantic model reduces an IoT application to three composable building blocks. Together they encode the full intent of an application: what to observe, when to act, and what to do — all expressed in domain expert language.

The framework intentionally avoids full ontology-based modeling (which forces experts into predefined categories) and language-based modeling (which adds cognitive overhead). Instead, AmI-DEU uses logic rules + pattern matching — the simplest formal structure that keeps domain experts in control while allowing the action machine to reason semantically.

Logic Rules EBNF Grammar (internal) Action Machine onto-AMI Matching
ELEMENT 01 Entity

Represents any context — either an input (sensor reading, user state, environment condition) or an output (result of an environment action). Entities can be linked to domain concepts to improve semantic matching.

ELEMENT 02 Condition (IF)

Evaluates context against defined rules and thresholds. Conditions trigger the application's response logic — they are the "when" of an intention: when heart rate exceeds X, when no movement is detected for Y minutes.

ELEMENT 03 Action

Describes how the smart environment should react: deliver a notification, change a device state, publish a new context, or activate a service. Actions are the "do" of an intention — and depend on available IoT resources.

ELEMENT 04 Concept

Represents domain-specific objects and their relationships — Person, House, Room, Device. Concepts augment application semantics and can connect to external data sources such as MQTT or APIs.

ELEMENT 05 Intention

The top-level object — a full representation of an anticipated end-user outcome. An intention subsumes user profile, domain knowledge, required resources, and all actions. It becomes the IaaC application after compilation.

OUTPUT IaaC Assignment

The compiled artifact: a self-described, enriched context object containing conditions, actions, and knowledge. Reduced in size but rich in meaning — ready for autonomous deployment on any ContextAA node.


04 / What Is an Intention?

From Everyday Goals to Compiled Applications

An intention in AmI-DEU is a high-level description of what a user wants to achieve — expressed in their own domain language. These are real examples of what domain experts can define, covering everything from simple reminders to full emergency protocols.

Behavior

"Play music to reduce stress when the system detects increased physiological indicators of anxiety in the user."

Daily Activity

"Support departure preparation: remind to prepare breakfast, get dressed, prepare lunch, and notify optimal departure time."

Scenario

"Ambient assisted living for an elderly person: cooking reminders such as add salt and turn off the stove."

Combination

"Remind me to rest 10 minutes each hour by displaying a non-intrusive alert on my computer and vibrating my chair softly."

Complex Protocol

"Emergency dispatch, control and supervision protocol combining multiple scenarios — medical, fire, security risk — and scenes such as chemical fire or security alert."


05 / Framework Components

What Makes Up AmI-DEU

The framework is composed of four tightly integrated components that together take an idea from a domain expert's head to a running autonomous agent in a smart environment.

COMPONENT 01 🎨 Visual IDE Node-RED-inspired · Flow Metaphor · EUD

A drag-and-drop interface with two workspaces: a Domains area for defining concepts and relationships, and an Applications area for building intention flows. Designed for End-User Development (EUD) — no programming background required.

Drag & DropFlow MetaphorEUDLow Cognitive Load
COMPONENT 02 ⚙️ Action Machine Semantic Compiler · Rule Matching · IaaC Generator

The brain of AmI-DEU. The action machine processes the expert's intention flow, matches rules and knowledge semantics using onto-AMI, and compiles a reduced but enriched IaaC application ready for ContextAA deployment.

Rule EngineSemantic MatchingIaaC CompilerEBNF
COMPONENT 03 📊 onto-AMI (UnOvi) Universal Ontology · Semantic Knowledge Base

AMI-Lab's universal ontology for smart environments. Provides the semantic backbone for concept matching and application compilation. Ensures that compiled intentions are interoperable across diverse IoT environments and device vocabularies.

OntologyInteroperabilityKnowledge BaseUnOvi
COMPONENT 04 🌐 ContextAA Integration Deployment Target · Context Matching · Multi-Environment

Compiled IaaC applications are deployed directly to ContextAA, AMI-Lab's distributed multi-agent platform. ContextAA handles context matching, agent coordination, and service continuity — whether in a smart home, hospital ward, or smart city.

ContextAAAutonomous AgentsContext MatchingMulti-Node
COMPONENT 05 🔌 Data Source Connectors MQTT · API · DomoSense · IntersCity

Domain concepts can be linked directly to live data sources: MQTT brokers (e.g., DomoSense gateway), REST APIs, or smart city data platforms (e.g., InterSCity). Context flows from physical sensors into the intention logic automatically.

MQTTREST APIDomoSenseLive Data
VALIDATED IN Testbed & Validation Smart Home · Smart City Simulator · Real IoT Devices

AmI-DEU has been validated end-to-end on a physical testbed using real IoT devices for smart home scenarios, and extended to a smart city simulator for scalability evaluation. Results show enhanced adaptation before and after deployment.

Smart Home TestbedSmart CityReal DevicesScalability

Want to build with AmI-DEU?

Whether you are a researcher, a domain expert, or a student — AmI-DEU is open for collaboration. Contact us or explore the full platform documentation.

Project Team

Bessam Abdulrazak

Professor, Department of Computer Science, Université de Sherbrooke

Director of the AMI-Lab

Victor Manuel Ponce Diaz

Ph.D. Student

Supervisor: Prof. Bessam Abdulrazak

Period: until September 2022

Publications

Ambient Intelligence Governance Review: From Service-Oriented to Self-Service

PeerJ Computer Science
Victor Ponce Bessam Abdulrazak

Context-Aware End-User Development Review

Applied Sciences
Victor Ponce Bessam Abdulrazak

Dynamic Domain Model for Micro Context-Aware Adaptation of Applications

2016 Intl IEEE Conferences on Ubiquitous Intelligence & Computing, Advanced and Trusted Computing, Scalable Computing and Communications, Cloud and Big Data Computing, Internet of People, and Smart World Congress
Victor Ponce Patrice Roy Bessam Abdulrazak

Intention as a Context: An Activity Intention Model for Adaptable Development of Applications in the Internet of Things

IEEE Access
Victor Ponce Bessam Abdulrazak