Fall response
Detect a fall event, escalate risk in the digital twin, propose check-in or alert actions, and only execute what the Safety Gate allows.
Safety-constrained adaptive architecture for robotic eldercare
An integrated architecture for AI-driven elderly support robots — with a deterministic Safety Constraint Gate between proposal and action.
The challenge
Ageing in place increases fall risk and care complexity, while care robotics remains fragmented across perception, reasoning, and human oversight.
The SAFE-CARE answer
Five coordinated layers — perception, digital twin and risk, constrained reasoning (LLM + Safety Gate), action/HRI, and care-network interfaces — designed so the LLM proposes and the gate disposes.
Module and simulation results are reported on the Evaluation page. No living-lab clinical validation study has been performed.
Illustrative form-factor atmosphere — not a product endorsement.
Use cases
Illustrative eldercare robotics scenarios SAFE-CARE is designed to support — always with the Safety Gate between proposal and action.
Medication reminder with Safety Gate
Schedule and voice reminders for known regimens; the gate blocks unsafe dosing suggestions and routes exceptions to human oversight.
Night check-in
Low-disturbance night rounds: presence and distress cues, quiet prompts, and care-network escalation when thresholds are crossed.
Care-home fleet
Shared robots across a residential care site — task queues, resident context via RCDT, and organization dashboards for staff.
Home ageing-in-place
A single companion form factor in a private home: daily routines, fall awareness, and family/care-org connectivity under constraint.
Research pilot
Instrumented evaluation deployments for universities and living labs — logging, metrics, and gated action policies without clinical claims.
Humanoid Robot
The physical platform
SAFE-CARE runs on humanoid and service-robot form factors — sensors, reach limits, and Safety Gate integration.
