From the Emergency Services edition of September 23, 2026
BEACON: USF multilingual agentic app for wildfire evacuation routing, checklists, and chat—prototype with qualitative Colorado tests
University of San Francisco MS DSAI program; Accenture collaborator · Academic prototype / wildfire public alerting and evacuation support · United States focus (LEP equity); qualitative experiments on Colorado 2025 fires
- Publisher
- arXiv
- Original publication
- 2026-09-03
- Source retrieved
- 2026-09-24
What happened
arXiv:2609.03301 (submitted 2026-09-03) presents BEACON, an iOS + FastAPI agentic service that ingests Watch Duty fire perimeters/orders/shelters, user GPS, and NOAA/HRRR weather to score Hourly Wildfire Potential, schedule XGBoost-driven reassessment, and generate polygon-avoidant evacuation routes via OpenRouteService—addressing commercial GPS failures that cross exclusion polygons. A multilingual LLM agent binds routing, personalized household checklists (12 binary profile questions), and UI language switching for limited-English-proficiency users, citing that over 80% of emergency messages are English-only despite ~26 million LEP residents. Preliminary qualitative tests on Colorado fires (Jul–Sep 2025) report reliable language detection, routes that avoided active polygons, first-turn context awareness, and graceful degradation (including call 911 when no safe route). Not a production agency deployment or measured outcome study.
Why it matters
Direct emergency-management/public-warning AI for wildfire evacuation equity—new relative to camera/drone detection coverage already archived. Paired with GAO multilingual planning findings in this edition for controls. Fresh preprint (2026-09-03) after last edition.
Evidence and measured results
Opened arXiv abstract and HTML/PDF body via WebFetch/API. Architecture, Watch Duty/NOAA inputs, HWP formula narrative, ORS polygon constraints, agent prompt blocks, and qualitative Colorado results taken from paper. Quantitative field outcomes not claimed.
Limitations and uncertainty
Preprint; qualitative demos only; depends on Watch Duty/third-party APIs; not an official NWS/FEMA product; Accenture co-author affiliation noted.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-23; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- LEP households often get English-only wildfire alerts and GPS routes that can cross fire perimeters.
- Stakeholders
- Emergency managers
- Public information officers
- GIS/IT
- Community LEP advocates
- Discovery questions
- What share of residents are LEP?
- Which alert channels are English-only?
- Do current apps respect fire polygons?
- Who owns unofficial guidance risk?
- Value hypothesis
- Multilingual agentic routing/checklists can improve inclusive preparedness if degraded modes escalate to 911 and agencies retain official authority.
- Potential engagement
- Limited geographic pilot with language QA and polygon-avoidance acceptance tests.
- Unsupported claims
- Official alerting replacement
- Measured lives saved
- Production-ready without privacy review
Pre-sales engineering
Role takeaway
- Fit
- Consumer/EMA-facing agentic app for evacuation guidance with GIS avoidance and multilingual UI.
- Architecture and integration
- iOS client, FastAPI on cloud, Watch Duty/NOAA/ORS/Mapbox, Qdrant memory, XGBoost cadence for HWP refresh.
- Prerequisites
- Perimeter/order feeds
- Shelter POI sources
- Push infra
- Language QA corpus
- Constraints
- Third-party API dependence
- Qualitative-only evidence
- Not an IPAWS originator
- Security
- Protect GPS trails and household medical/mobility flags; define retention and model-training bans.
- Proof of value
- Polygon avoidance vs commercial GPS
- Language consistency audits
- 911 fallback drills
Delivery
Role takeaway
- Work
- Stand up pilot feeds, recruit LEP testers, run degradation drills, and coordinate with official alert owners.
- Dependencies
- Watch Duty/NOAA access
- EMA counsel/privacy
- Community translators
- Ownership
- EMA PIO/ops own public messaging boundaries; vendor/university owns app ops under contract.
- Skills and adoption
- Train staff not to over-endorse unofficial routes; teach testers when to call 911.
- Governance checkpoints
- Privacy impact assessment
- Content disclaimer approval
- Failover drill sign-off
- Acceptance criteria
- Avoidance tests passed
- Language switch verified
- No-safe-route to 911 path proven
- Risks
- Users treat app as official alert
- Stale perimeter data
- Chat hallucinations
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Mobile client + cloud agent with GIS avoidance routing, vector memory, and dynamic language propagation across UI/chat/checklist. Agency adaptations need authoritative perimeter feeds, shelter APIs, and offline/degraded modes.
Governance
Who approves, reviews and stays accountable for outcomes?
Prototype must not become unofficial alerting authority; require agency content control, 911 escalation rules, and clear non-clinical scope (paper states no clinical advice).
Security and privacy
What data, permissions and controls need testing?
Stores GPS, household vulnerability profile, and chat memory (Qdrant); needs privacy impact assessment before public-sector pilots.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Explicitly targets LEP inclusion via UI/chat language switching and personalized household checklists including mobility needs.
Procurement
What should contracts, pricing and exit terms secure?
University/prototype status favors research MOU or staged RFP with data-rights and IPAWS non-conflict clauses—not sole-source alerting procurement.
Operating model
Which teams own the service once it runs?
Adds a parallel public guidance channel that must stay subordinate to official WEA/EAS and 911.
Publication history
- 2026-09-23Emergency Services · Issue 185 resources
Stable resource ID: arxiv-beacon-multilingual-wildfire-evacuation-agent-20260903