AI-powered SAR mission planning with lost person behavior modeling, probability-based sector allocation, and real-time Bayesian updates. Based on ISRID statistics and Koester's research, the system predicts subject travel patterns, optimizes team assignments, and tracks search progress through POA, POD, and POS metrics — completely offline.
The Lost Person Behavior Service implements statistical models from the International Search & Rescue Incident Database (ISRID) and Robert Koester's research. Each subject scenario type maps to empirical data on travel distances, terrain preferences, and time-to-find statistics.
| Scenario Type | 50% Found Within | 95% Found Within | Typical Behavior |
|---|---|---|---|
| Lost Hiker | 3.2 km | 16.0 km | Follows trails, seeks high ground |
| Dementia | 1.9 km | 12.8 km | Linear travel until obstructed |
| Despondent | 1.6 km | 12.0 km | Limited movement, may hide |
| Child (1-3) | 0.4 km | 2.4 km | Very limited range |
| Child (7-12) | 1.6 km | 9.6 km | May attempt self-rescue |
An 82-year-old with dementia wanders from a care facility. The SAR team creates a mission profile selecting "Dementia" as the scenario type. The system immediately displays the statistical profile: 50% found within 1.9km, tendency to walk in straight lines, attraction to roads but possible hiding from vehicles. The AI recommends helicopter thermal imaging for linear feature coverage and roadblock teams. The initial search sectors are sized at 2km radius rather than the default 5km, focusing resources on high-probability areas.
SAR operations use standard probability metrics that the system tracks and updates automatically:
As teams complete sectors and report POD achieved, the system runs Bayesian updates to redistribute residual POA to remaining sectors. This ensures search priorities always reflect the latest information.
The on-device AI analyzes subject profiles, clue patterns, and team capabilities to generate recommendations:
The AI generates situation-specific assessments based on subject age, medical conditions, clothing, and scenario type. It identifies risk factors, likely behaviors, and recommended resources.
Multiple clues are analyzed for movement patterns, direction consistency, and convergence zones. The AI can detect circulatory patterns (confusion), linear travel (dementia), or fast movement (running/vehicle).
Given a list of available teams with capabilities (K9, technical rescue, swiftwater, etc.) and sector requirements, the system calculates optimal assignments using a multi-factor scoring algorithm that considers proximity, capability match, fatigue levels, and sector priority.
The Search Pattern Service generates optimal patterns based on scenario type and available resources:
| Pattern | Best For | Baseline POD | Coverage Rate |
|---|---|---|---|
| Expanding Square | Hasty search around LKP | 35% | 2.0 km²/team-hour |
| Sector Search | High-probability confinement | 85% | 0.5 km²/team-hour |
| Parallel Track | Large area systematic | 75% | 1.0 km²/team-hour |
| Grid Type I | General area coverage | 60% | 1.0 km²/team-hour |
| Grid Type III | High-certainty areas | 90% | 0.25 km²/team-hour |
Integrated weather analysis evaluates conditions for team safety, subject survival probability, and search methodology adjustments. The system calculates survival time modifiers based on temperature, wind, precipitation, and subject profile.
A search initiates at 2 AM with temperatures at -10°C for an overdue hiker. The weather impact assessment automatically flags extreme hypothermia risk for both subject and teams, recommends extreme cold weather gear, reduces estimated subject survival time by 50%, and prioritizes high-speed search methods (ATV, helicopter) over slow grid searches. The AI recommends immediate water source checks since subjects often follow watercourses in cold conditions.
The briefing generator compiles subject profiles, search progress, clue analysis, and AI assessments into comprehensive mission briefings suitable for:
The full SAR workflow is accessible through natural language via dedicated agent tools:
Example: "Create a new SAR mission for a missing 6-year-old last seen at these coordinates. The child was wearing a red jacket and has no medical conditions. Analyze the likely behavior pattern and recommend initial search sectors."