Smart Spaces Architecture & Specs

Aeion moves beyond simple reactive thresholds, detecting inefficiencies like "Ghost Heating" (HVAC running while occupancy is zero) and grading the entire facility A-F.

Efficiency Scoring Model:

Every building gets an overall score (0-100) mapped to a letter grade:

• A: 90+ (top-tier efficiency)
• B: 80+ (good, minor improvements possible)
• C: 65+ (average, significant savings available)
• D: 50+ (below average, action needed)
• F: under 50 (critical, immediate attention required)

Alongside the overall grade, each building gets a category breakdown — HVAC, lighting, equipment, and plug loads — each with its own score, letter grade, benchmark comparison against similar buildings, and a dollar-value savings estimate if optimized, plus a confidence rating.

Ghost Heating Detection:

Aeion continuously compares two signals per zone: how occupied it's been (from motion, badge, or camera sensors) and how hard the HVAC has been running. When a zone sits essentially empty for an extended period while its HVAC keeps working, that's flagged as ghost heating — rated by severity, with the wasted energy translated into an estimated annual dollar cost and a specific recommended fix, such as reducing the setpoint automatically 15 minutes after the last occupant leaves.

Each building's overall grade blends its HVAC, lighting, equipment, and plug-load scores — weighted 40% / 25% / 20% / 15% respectively — into one number, rounded to a letter grade.

AI Recommendations:

Every recommendation includes a title, category, estimated annual savings, effort level (low/medium/high), a priority ranking, and a confidence score — sorted so the highest-value, lowest-effort fixes surface first. Examples:

• "Reduce HVAC setpoint 15 minutes after last occupancy detected" — low effort, $12K/year
• "Schedule equipment shutdown during weekend unoccupied hours" — low effort, $4K/year
• "Add occupancy-based lighting zones in open floor plan" — medium effort, $8K/year
• "Deploy predictive maintenance on the AHU showing a vibration anomaly" — high effort, $15K/year

Weekly digest: Every Monday, facilities managers get an automated summary: total consumption vs. the prior week, cost trend, top savings opportunities, how much active automation rules saved, and the building's updated efficiency score.

Bidirectional sync between facility room bookings and Google Calendar/Outlook, with resource-level conflict detection.

Calendar Integration:

External meetings booked in Google Calendar or Outlook automatically create a matching room hold in Aeion; bookings made directly in Aeion send calendar invites to attendees. It stays in sync in both directions.

Conflict detection:

Every booking request is checked against existing confirmed, checked-in, and pending-approval bookings for the same resource (room, desk, or piece of equipment) and time window — covering double-booking across rooms and scheduling clashes over shared equipment. When a request overlaps an existing booking, Aeion returns the specific conflicting bookings instead of silently allowing a double-booking through.

Room & resource booking: Real-time availability across rooms, desks, and equipment, with support for recurrence, catering orders, extra setup/teardown buffers, and automatic calendar invites.

High-performance telemetry ingestion with a buffer + batch flush pattern, append-only storage, and per-topic filtering.

Ingestion Architecture:

Every connected sensor publishes to a topic scoped to its building, zone, sensor type, and metric. Aeion buffers incoming readings and flushes them to storage every second, or every 100 messages — whichever comes first — so ingestion keeps pace under heavy load without per-message database writes, and stays efficient when things are quiet.

Topic Schema:

``` aeion/{buildingId}/{zoneId}/{sensorType}/{metric}

aeion/building-a/zone-3/temperature/current → 72.4°F aeion/building-a/zone-3/occupancy/person_count → 12 aeion/building-a/main-panel/electrical/power → 8,420W aeion/building-a/hvac/ahu-1/runtime → 87% ```

Append-Only Enforcement:

Once written, a telemetry reading can never be updated or deleted. That matters for two reasons: your ESG/sustainability reporting data stays tamper-proof and audit-ready (including for ISO 50001 energy-management compliance), and the historical data used to train AI models can't be corrupted after the fact.

Query any time range, zone, or metric on demand — results ordered chronologically, capped at 10,000 rows per request for fast response times.

Floor and zone data power both the digital twin and indoor wayfinding.

Upload an SVG or DWG floor plan and Aeion automatically extracts door centroids, wall geometries, and zone labels (id, name, type, boundary polygon) into a GeoJSON spatial index on the floor record — no manual tracing. A floor can also carry a GLB/GLTF digital-twin model for 3D visualization. Indoor wayfinding runs on a graph of nodes (elevators, stairs, corridors, resources — each with x/y/z coordinates) and edges connecting them, including paired nodes across floors so routes can change levels via elevator or stairs.

Indoor Wayfinding:

A\* pathfinding calculates the optimal route between any two points on this graph — accounting for doorways, stairs, and elevators — with two modes:

• Standard — optimal path across the full graph
• Accessible — restricted to wheelchair-accessible edges, using elevator (not stairs) transitions between floors

A route result returns the ordered waypoints, total distance, estimated walking time, and which floors it crosses. If no navigation graph is configured for a building, Aeion falls back to an honest straight-line estimate rather than fabricating a route.

A real-time event backbone connects telemetry ingestion to every downstream consumer — dashboards, automations, and other Aeion modules.

Every telemetry change is published as soon as it happens. Threshold crossings (like a zone's temperature exceeding a safe limit) and occupancy changes (a zone going from occupied to empty) broadcast instantly to anything subscribed — including the automation rule engine, which can react immediately, for example turning off zone lighting the moment it empties out.

Cross-module coordination: Because this event backbone is shared across the whole platform, other modules react to Smart Spaces events natively — no custom integration required. A Sentinel security lockdown can isolate floors via elevator control; an occupancy change can trigger Sentinel's camera tracking; an emergency override can trigger HVAC emergency ventilation.

Live Dashboards:

NOC dashboards and 3D digital twins subscribe over WebSocket to live occupancy, energy, and alert channels — authenticated and scoped to your tenant:

typescript ws.send(JSON.stringify({ action: "subscribe", channel: "occupancy:zone_3" })); ws.send(JSON.stringify({ action: "subscribe", channel: "energy:building_a" })); ws.send(JSON.stringify({ action: "subscribe", channel: "alerts:all" }));

Updates reach the browser in well under a second, with no polling and no stale data — a zone occupancy change is typically visible in the dashboard within a quarter-second.

Multi-protocol IoT device abstraction layer supporting MQTT, BACnet/IP, Modbus RTU/TCP, KNX, EnOcean, Z-Wave, Zigbee, Thread, Matter, Wi-Fi, Bluetooth LE, LoRaWAN, OPC-UA, DALI, and a custom HTTP/REST adapter — all behind a single, unified command interface. Vendor BMS/lighting platforms (Siemens, Honeywell, Johnson Controls, Schneider Electric, Crestron, Lutron, Philips Hue, Tuya) come along for free wherever their hardware speaks one of these standard protocols.

Every device command — read a value, write a value, or subscribe to changes — goes through the same interface regardless of which protocol the underlying hardware speaks. That means a temperature setpoint change to a legacy BACnet controller and a lighting scene change to a Zigbee bulb use identical application logic; only the adapter underneath differs.

BACnet/IP Integration:

For legacy commercial building HVAC systems, Aeion bridges BACnet devices directly, supporting property reads and property writes. For example, setting a zone's temperature setpoint to 72°F is a single write call that respects BACnet's standard priority-array semantics; reading the current temperature back returns the live value with its timestamp.