[SESSION_127] FallDetector: Implement Multi-Person Simultaneous Falls (Phase 23)

Feature: Multi-Person Simultaneous Fall Detection

Component: FallDetector (/nodes/behavior_nodes/fall_detector/)
Severity: 🔴 CRITICAL BLOCKER
Session: SESSION 127 (Definition of Done)
Current Status: Only processes one person per frame
Target Phase: Phase 23

Context

Fall detection in crowd environments requires detecting multiple simultaneous falls. Currently, the node processes only one person per frame, missing crowd accident scenarios.

Requirements

• Batch process multiple persons per frame
• Detect simultaneous falls from multiple people
• Aggregate and correlate events
• Maintain independent state per person
• Generate separate events for each person

Impact

Without this feature:

• False negatives in crowd scenarios
• Missed accident detection (critical for liability)
• Unfit for retail/public space deployments
• Limited to single-person retail environments

Effort Estimate

• Person batching: 2 hours
• Event aggregation: 1 hour
• Testing with crowds: 1-2 hours
• Total: 3-4 hours

Technical Approach

• Parallel processing of person tracklets
• Per-person state tracking dictionary
• Event batching and deduplication
• Concurrent detection without race conditions

---

Generated by SESSION 127 FASE 2.7