Spanda

Ambient Intelligence

Context-aware, safety-verified orchestration for intelligent environments — part of the Smart Spaces & Ambient Intelligence blueprint.

Status: Experimental (scaffold)

Ambient intelligence here means: sense context → verify readiness → orchestrate missions → produce evidence — not opaque black-box automation.

Definition

Term	Spanda interpretation
Ambient	Environment-wide signals (occupancy, AQ, energy, presence)
Intelligence	Verified missions + optional AI agents within capability bounds
Context	Entity graph: who is where, what devices are healthy, what mode applies

Spanda adds verification and trust layers that typical ambient-AI stacks omit.

Context stack

Sensors & wearables
  → Entity graph (building, zone, occupant, device)
  → Readiness engine (can we act safely?)
  → Mission planner (.sd programs + optional AI agents)
  → Providers (Matter, BACnet, MQTT, robots)
  → Assurance evidence + Control Center

Context signals

Signal	Sources	Missions influenced
Occupancy	PIR, mmWave, BLE, cameras	HVAC setback, lighting, cleaning
Presence	Wearables, mobile app	Health alerts, personalized climate
Air quality	CO₂, VOC, particulate	Ventilation boost, notify operator
Energy price	Utility API, `spanda-energy`	Demand response, EV defer
Safety	Smoke, CO, leak	Emergency override all comfort missions
Trust	Lock tamper, camera attestation	Block access missions

Adaptive workflows

Occupancy-driven climate

Office zone unoccupied for 30 minutes → readiness checks BACnet path → setback HVAC → record assurance snapshot.

Example: examples/solutions/smart-spaces/smart-office/occupancy_climate.sd

Presence-aware lighting

Occupant enters room → verify gateway + dimmer health → raise lights → log access in assurance bundle.

Health-adjacent ambient (opt-in)

Wearable fall signal → verify responder availability → emergency mission → Connected Healthcare bridge.

Example: examples/solutions/smart-spaces/hospital-at-home/patient_monitoring.sd

AI agents

AI agents (via spanda-openai, spanda-onnx, or custom packages) may propose mission parameter changes. They do not bypass:

requires capabilities verification
Readiness score thresholds
Human approval for lockdown, medical, or life-safety overrides
kill_switch and health_policy enforcement

Privacy

Occupant and health dimensions follow human-readiness.md opt-in policy.
Camera and microphone streams require explicit capability grants in device tree.
Assurance bundles redact PII per deployment spanda.security.toml.

Simulation

Test ambient scenarios without live devices:

spanda sim examples/solutions/smart-spaces/smart-office/occupancy_climate.sd

Fault injection: sudden occupancy spike during night mode, AQ threshold breach, gateway loss mid-mission.

building-automation.md — BMS and device control
human-interaction.md — Operator and occupant entities
solutions/spatial-computing.md — AR / wearable context

This site is open source. Improve this page.