How RescueAI works
Most emergency systems optimise the responder side. Most emergency guidance focuses on the individual. RescueAI connects both. For the person in danger, RescueAI reduces cognitive load by providing one simple action at a time. For responders, RescueAI turns the incident into a structured, simulated environment where hazards, routes, people and rescue priorities can be examined together.
βDon't overwhelm. Guide.β
Architecture
- 01User reports the emergency
- 02RescueAI interprets the situation
- 03Structured state is created
- 04Verified protocol engine chooses guidance
- 05Command Center receives the incident state
- 06Simulation models routes and hazards
- 07Pathfinding recommends safer routes
- 08Impact metrics are calculated
USER INPUT β SITUATION INTERPRETER (AI, schema-validated Β· keyword fallback) β STRUCTURED STATE β VERIFIED PROTOCOL ENGINE (deterministic branching) β CURRENT SAFE ACTION β USER β INCIDENT MODEL β SIMULATION β PATHFINDING β IMPACT METRICS
Technical features
AI situation interpretation
Free text β validated structured state, with deterministic fallback.
Deterministic protocol engine
Safety-critical branching comes from authored protocol data, never a model.
Voice guidance
Browser speech synthesis reads every step aloud, with a stop control.
Grid-based simulation
Multi-floor building, deterministic fire spread, congestion and queueing.
Pathfinding
Weighted A* over the grid, treating hazard cells as impassable and risk cells as costly.
Dynamic route recalculation
Routes are re-planned whenever hazard spread blocks the next cell.
Prototype rescue prioritisation
Transparent scoring with a visible explanation per person.
Before / after evaluation
Two headless runs on identical inputs produce the impact comparison.
Safety & transparency
- β’RescueAI is a hackathon prototype.
- β’It is not a certified emergency-response system.
- β’It does not replace emergency services.
- β’It does not replace trained medical personnel.
- β’Safety-critical guidance must come from verified emergency protocols.
- β’Simulation results are not real-world performance claims.
- β’AI output is constrained and validated against a strict schema.
- β’If AI is unavailable, deterministic fallback mode is used.
RescueAI is a prototype for research and emergency-response interface design. It is not a substitute for trained emergency responders, professional medical care, or official emergency guidance.
This dials your device's emergency number. RescueAI never dispatches police, fire or ambulance services itself.