SunogEskinita: Alleyway fire spread and hose-relay pathfinder for dense urban sitios

Problem Statement

Residential fires in dense urban communities such as Tondo, BASECO,Addition Hills, Mandaue, and coastal Davao destroy hundreds of homes in a single hour. Houses built with light timber, plywood, and corrugated iron roofs sit less than half a meter apart, allowing radiant heat and flame impingement to jump across structures in 3 to 5 minutes. Standard map apps stop at the main road and show dense sitios as blank gray polygons. When a fire starts inside an eskinita labyrinth, residents fleeing in smoke often run into locked gates or dead-end alleys, while Bureau of Fire Protection trucks cannot fit inside 1-meter passages and lose precious minutes figuring out how many 50-foot hoses are needed to reach the seat of the fire.

Proposed Solution

SunogEskinita is a community-mapped fire escape and hose-relay navigation system built for high-density urban barangays. Before an emergency, barangay fire volunteers and youth council members walk each eskinita recording a 2-minute smartphone video while tagging standpipes, fire hydrants, overhead wire tangles, LPG stores, and dead-end gates. Computer vision extracts alley width, clearance height, and wall/roof construction materials, building a walkable micro-graph stored in PostGIS.

When a resident or tanod reports a fire pin via SMS, Messenger, or a one-tap barangay alarm, SunogEskinita pulls real-time wind speed and direction and runs a fast cellular-automaton fire and smoke spread simulation calibrated on full-scale informal settlement fire experiments. Within 10 seconds, it sends two outputs:

1. For residents: Plain Tagalog or Bisaya SMS and Messenger alerts naming the exact eskinita landmarks to take so families move upwind and perpendicular to the flame front, avoiding blocked alleys.
2. For barangay fire volunteers and BFP responders: A hose-lay pathfinder showing the exact road parking drop-point, the shortest alley path, and the exact count of 50-foot hose lengths required to reach the burning cluster.

Target Market

• Primary users: Residents, barangay tanods, and community fire auxiliary brigades in high-density urban poor communities across Metro Manila, Metro Cebu, Metro Davao, Iloilo, and Bacolod.
• Public-sector users: Bureau of Fire Protection city stations, city DRRM offices, and barangay councils utilizing their 5 percent Local DRRM Fund.
• Civil society partners: Philippine Red Cross community-based disaster programs, urban poor federations, and volunteer fire brigades.

How AI Is Used

• Roof and alley vulnerability segmentation: Computer vision segments satellite tiles and smartphone walkthrough frames into light timber, concrete firewall, and metal sheet roofs while estimating alley width and overhead wire obstructions.
• Physics-informed cellular automaton fire spread model: Predicts 5-, 10-, and 15-minute ignition envelopes using thermal radiation and wind-driven flame tilt rules derived from informal-settlement burn experiments.
• Dynamic hazard-avoidant graph routing: Recomputes Dijkstra/A* pedestrian evacuation paths on the eskinita micro-graph every 60 seconds, penalizing alleys downwind of smoke plumes or adjacent to LPG hazards.
• Multilingual landmark direction writer: Converts graph nodes into familiar neighborhood directions (“Lumabas sa Eskinita 3 dumaan kina Aling Nena papuntang basketball court; huwag dumaan sa Eskinita 1 dahil mausok”).

Community & Ecosystem Integration

• Barangay Volunteer Fire Brigades: Gives local volunteers who arrive in the first 3 minutes a concrete hose-relay map before BFP pumper trucks arrive.
• Zero-app resident access: Works over plain SMS, Messenger, and barangay public-address megaphones so families do not need to install an app during a panic.
• OpenStreetMap Philippines & Humanitarian Mapping: Exports verified micro-alley geometries back to open community maps for ambulance and flood rescue use.

Demo Scenario (3 minutes for judges)

1. Show a satellite view of a dense urban barangay in Manila where Google Maps shows zero internal roads, then toggle on the SunogEskinita micro-alley graph showing 42 mapped eskinitas, 3 dead ends, and 2 street hydrants.
2. Click a dwelling in Purok 3 to trigger a fire report with a 15 km/h northeast wind.
3. The fire-spread model projects the 5-minute and 10-minute ignition cones in red and orange. Two alleys immediately turn black on the map as smoke traps.
4. A judge’s phone receives a Tagalog SMS giving the exact safe eskinita exit route toward the barangay covered court, while the responder screen displays: “Park pumper at Kalye Mabini corner; lay 6 lengths (300 ft) of 1.5-inch hose via Eskinita 4.”
5. Simulate an alley blockage reported by a tanod; the graph reroutes all evacuation arrows and hose paths in under 2 seconds.

30-Hour Build Plan

Suggested stack: Python FastAPI, Ultralytics YOLOv8, NetworkX, PostgreSQL/PostGIS, Next.js + MapLibre GL, Gemini or open LLM for landmark instructions, SMS gateway. Public data / APIs: OpenStreetMap Philippines building footprints, Google Open Buildings, PAGASA real-time wind telemetry, BFP public fire incident statistics, Project NOAH urban density layers.

Research & Literature References

1. Antonio Cicione, Richard S. Walls, and Charles Kahanji, “Experimental study of fire spread between multiple full scale informal settlement dwellings,” Fire Safety Journal, Vol. 105, pp. 19-27, 2019. DOI: 10.1016/j.firesaf.2019.02.001
   • How SunogEskinita uses this study: Uses the empirical heat-flux, flashover timing, and dwelling-to-dwelling ignition thresholds measured in full-scale informal settlement dwellings to calibrate the 3-to-5-minute spread rate between timber and metal-clad shanties.
2. Antonio Cicione, Colleen Wade, Michael Spearpoint, Lesley Gibson, Richard Walls, and David Rush, “A preliminary investigation to develop a semi-probabilistic model of informal settlement fire spread using B-RISK,” Fire Safety Journal, Vol. 120, 103115, 2021. DOI: 10.1016/j.firesaf.2020.103115
   • How SunogEskinita uses this study: Adapts the semi-probabilistic radiation and flame-impingement rules into a fast graph-based cellular automaton that runs in seconds during an active barangay fire.

Risks & Privacy Notes

• Life-safety fallback: Automated SMS directions always include the barangay’s pre-designated primary assembly points and must align with the Barangay Fire Brigade’s physical signage.
• Resident privacy in walkthrough videos: The alley-mapping pipeline automatically blurs faces, house numbers, and license plates before storing alley clearance geometries, complying with the Data Privacy Act of 2012.
• No false precision: Fire spread polygons are shown as probabilistic risk zones rather than guaranteed boundaries, prompting early evacuation of the entire affected block.
• Strict 30-hour feasibility constraint: Routing requires pre-walked video surveys of every informal settlement alley before a fire occurs; unmapped barangays cannot use the hose-path graph during an emergency.

AI Use Disclosure (draft)

SunogEskinita uses YOLOv8 to classify roof materials and alley obstructions, a physics-informed cellular automaton to model fire spread, and an LLM to turn graph routes into landmark-based Tagalog evacuation messages. AI coding tools assisted in writing the NetworkX routing module and MapLibre visualization layers. All fire spread parameters were verified against peer-reviewed fire engineering studies.

Rubric Scoring