Flight Alerter & Air Traffic Visualization Migration to Uncrew

Andi Lamprecht ·2026-05-11· 13 min read· Draft

Goal: Retire the Clojure-based UTM Themis flight-alerter and airspace visualization by migrating all features into the Uncrew platform (Go backend + Apollo React frontend).

1. Executive Summary

DroneUp currently runs a Clojure-based UTM Themis stack for Detect-and-Avoid (DAA) functionality:

• utm-themis-daa — Ingests raw ADS-B data (ASTERIX CAT021) from ground receivers, parses, geo-filters, stores plots in PostgreSQL, and publishes to Redis.
• utm-themis-api — REST + WebSocket API that streams enriched plots (with intrusion markers) to subscribers, manages flight plans, and computes OCDV/OCAV intrusion detection.
• utm-themis-grpc — gRPC streaming service that reads from Redis journals, fetches full plots from PostgreSQL, runs DAA intrusion detection, and streams to consumers.
• utm-themis-spaces — Clojure/ClojureScript full-stack web app providing the Airspace Manager UI with real-time 3D map, DAA alert visualization, audio alerts, flight plan management, and operator compliance reports.
• utm-clojure-kit (“duck”) — Shared library containing CAT021 parser, DAA algorithms (OCDV/OCAV), geometry functions, and infrastructure abstractions.

This PRD covers migrating everything downstream of ADS-B ingestion (i.e., the flight-alerter / conflict detection logic, real-time streaming, and frontend visualization). The CAT021 ingestion pipeline (utm-themis-daa) is explicitly out of scope since Uncrew already handles air traffic data ingestion (see Airdex and future onboard ADS-B ingestion).

2. Scope

In Scope

Out of Scope

3. Current System Architecture (What We’re Replacing)

ADS-B Receiver (uAvionix)
     | UDP (ASTERIX CAT021)
     v
utm-themis-daa (Clojure) -- ALREADY REPLACED BY UNCREW INGESTION
     | Redis journal + PostgreSQL
     v
utm-themis-grpc (Clojure)
     | gRPC stream (enriched plots with OCDV/OCAV markers)
     v
utm-themis-api (Clojure)
     | WebSocket push
     v
utm-themis-spaces (ClojureScript)
     | Browser: Mapbox GL + Threebox 3D
     v
Operator sees: aircraft, drones, alerts, airspace

Target Architecture (What We’re Building)

Uncrew ADS-B Ingestion (existing)
     | Internal data store (plots)
     v
Uncrew Conflict Detection Service (NEW Go service)
     | Reads plots + drone telemetry
     | Computes OCDV/OCAV intrusions
     | Streams alerts + enriched plots via gRPC
     v
Uncrew Apollo Frontend (ENHANCED React app)
     | gRPC-Web / Connect
     | Mapbox GL + deck.gl
     v
Operator sees: aircraft, drones, alerts, airspace

4. Backend: Conflict Detection Service (Go)

4.1 Service Purpose

A new Go microservice within the Uncrew ecosystem that continuously evaluates proximity between active drone flights and nearby manned aircraft (ADS-B plots), producing real-time intrusion alerts.

4.2 Core Algorithm: Two-Volume DAA Model

The service must implement the same OCDV/OCAV intrusion detection algorithm currently in utm-clojure-kit/daac.cljc, with updated timing constants per VT MAAP Chapter 1 Report (March 2025) section 4.5.5 recommendations:

OCDV (Ownship Centric Detection Volume)

• Shape: Cylinder centered on the drone (ownship)
• Horizontal radius: 2,778 meters (~1.5 statute miles)
• Vertical half-height: 76.2 meters (250 feet)
• Purpose: Outer detection zone — “be aware, monitor”

OCAV (Ownship Centric Alert Volume)

• Shape: Cylinder centered on the drone (ownship)
• Horizontal radius: Dynamic, depends on intruder speed and relative geometry
• Vertical half-height: 76.2 meters (250 feet)
• Purpose: Inner alert zone — “take evasive action NOW”

OCAV Radius Calculation

OCAV_radius = well_clear_radius + (intruder_airspeed * time_to_avoid)
time_to_avoid = intruder_position_uncertainty
              + pilot_delay
              + gcs_latency
              + time_to_initiate_maneuver
              + time_to_well_clear(ownship, intruder)
              + deceleration_buffer
              + safety_margin

Algorithm Constants

Updated per VT MAAP Chapter 1 Report (March 2025) section 4.5.5 recommendations. Previous values shown for traceability.

Intrusion Detection Logic

intrusion?(ownship, radius, height, plot):
  distance = haversine_distance(ownship, plot)
  delta_altitude = abs(plot.altitude - ownship.altitude)
  // If plot altitude is unknown, assume same altitude (worst case)
  if plot.altitude == nil:
    delta_altitude = 0
  return delta_altitude <= height AND distance <= radius

Alert States

4.3 Required Data Inputs

4.4 Required Data Outputs

4.5 gRPC Service Definition (Proposed)

service ConflictDetectionService {
  // Stream enriched plot data with intrusion status
  rpc StreamPlots(StreamPlotsRequest) returns (stream PlotUpdate);
  // Stream intrusion alerts for active flights
  rpc StreamAlerts(StreamAlertsRequest) returns (stream IntrusionAlert);
  // Get historical intrusion events for a flight
  rpc GetFlightIntrusions(GetFlightIntrusionsRequest) returns (FlightIntrusionsResponse);
  // Get current alert state for all active flights
  rpc GetActiveAlerts(GetActiveAlertsRequest) returns (ActiveAlertsResponse);
}

message PlotUpdate {
  string aircraft_id = 1;
  double latitude = 2;
  double longitude = 3;
  double altitude_meters = 4;   // WGS-84 meters
  double airspeed_mps = 5;
  double heading = 6;
  google.protobuf.Timestamp as_of = 7;
  AircraftType aircraft_type = 8;  // FIXED_WING, HELICOPTER, UAV, UNKNOWN
  IntrusionStatus intrusion_status = 9;  // CLEAR, OCDV, OCAV
  string nearest_mission_id = 10;  // Which drone is this closest to
}

message IntrusionAlert {
  string alert_id = 1;
  string mission_id = 2;         // The drone flight
  string aircraft_id = 3;        // The intruding aircraft
  AlertType alert_type = 4;      // OCDV_ENTER, OCDV_EXIT, OCAV_ENTER, OCAV_EXIT
  double distance_meters = 5;
  double ocav_radius_meters = 6;
  google.protobuf.Timestamp timestamp = 7;
}

enum IntrusionStatus {
  CLEAR = 0;
  OCDV = 1;
  OCAV = 2;
}

enum AlertType {
  OCDV_ENTER = 0;
  OCDV_EXIT = 1;
  OCAV_ENTER = 2;
  OCAV_EXIT = 3;
}

enum AircraftType {
  UNKNOWN = 0;
  FIXED_WING = 1;
  HELICOPTER = 2;
  UAV = 3;
}

4.6 Geospatial Filtering

The service must filter plots by geographic relevance before running DAA detection:

1. Site proximity filter: Only process plots within 10 miles (16,093.4 meters) of any active site
2. Haversine distance: Use great-circle distance for all proximity calculations
3. Site locations should be database-driven (not hardcoded like the current Clojure implementation)

4.7 Staleness & Cleanup

• Plot staleness: Plots older than 5 seconds should be considered stale and removed from active tracking
• Telemetry staleness: Drone telemetry older than 5 seconds should be considered stale

4.8 Performance Requirements

• Latency: Plot-to-alert latency must be < 500ms
• Throughput: Must handle 1000+ aircraft plots/second across all hubs simultaneously
• Concurrent flights: Must support 50+ simultaneous active drone flights

5. Frontend: Apollo Air Traffic & Alert Features

5.1 Technology Context

The uncrew-apollo-frontend is a React/TypeScript application using:

• Map: Mapbox GL JS v2.14+ with react-map-gl, deck.gl v9
• State: Zustand
• Data: gRPC-Web via @connectrpc/connect, TanStack Query
• UI: MUI Material v7, @droneup/uncrew-react-design-system
• Telemetry: Existing drone telemetry subscription via gRPC streaming
• Geometry: @turf/turf for geospatial operations
• Auth: Frontegg (@frontegg/react)

5.2 Feature: Air Traffic (ADS-B Plot) Visualization

Priority: P0 — Must Have

Display nearby manned aircraft positions on the existing map in real-time.

Requirements

Data Source

• gRPC stream from traffic source
• Filtered by geographic bounds visible on the map

Visual Specification (Matching Current Spaces Behavior)

5.3 Feature: DAA Alert Visualization

Priority: P0 — Must Have

Visualize OCDV and OCAV volumes around active drones, showing when manned aircraft are intruding.

Requirements

5.4 Feature: Audio Alerts

Priority: P0 — Must Have

Provide audible warnings when aircraft enter or exit detection/alert volumes.

Requirements

Audio Files

Reuse existing audio assets from utm-themis-spaces:

• enter.mp3 — attention alert
• warning.mp3 — critical alert (2-second duration)
• exit.mp3 — all-clear

5.5 Feature: Alert Notification Panel

Priority: P0 — Must Have

A sidebar or panel showing active intrusion alerts and their status.

Requirements

5.8 Feature: Operator Compliance Reports

Priority: P1 — Should Have

Post-flight reports showing intrusion events for regulatory compliance.

Requirements

6. New Zustand Stores Required

7. New gRPC Clients Required

8. Migration Considerations

8.1 Data Continuity

• No data migration needed — this is a runtime service, not a data store.
• Historical intrusion data from the old system is retained in the Themis PostgreSQL database for regulatory compliance; it does not need to be migrated but should remain accessible.

8.2 Parallel Operation

• The new system should run in parallel with the old Themis Spaces UI during transition.
• Operators should be able to use either system until confidence is established.

8.3 Known Issues to Fix During Migration

These issues exist in the current Clojure implementation and should be addressed:

8.4 What Can Be Retired After Migration

Once all features are live in Apollo and verified:

9. Acceptance Criteria Summary

P0 (Must Have — Required for launch)

• Conflict Detection Service running in Go, computing OCDV/OCAV in real-time
• ADS-B aircraft visible on Apollo map with real-time position updates
• Aircraft color-coded by intrusion status (clear/OCDV/OCAV)
• OCAV alert circles rendered around drones with nearby aircraft
• Audio alerts on intrusion state changes
• Alert notification panel showing active intrusions
• Plot staleness cleanup (>5s auto-removal)
• DAA algorithm constants match updated values from VT MAAP Chapter 1 section 4.5.5 (safety-critical)

P1 (Should Have — Required for Themis retirement)

• Dynamic hub location management

P2 (Nice to Have — Improvements over current system)

• Aircraft type detection from ADS-B emitter category (instead of hardcoded tail numbers)

10. Dependencies & Risks

11. References

• Deep Analysis of UTM Themis DAA
• Deep Analysis of UTM Themis Spaces
• utm-themis-daa repository
• utm-themis-spaces repository
• utm-clojure-kit (duck) — DAA algorithm source (src/cljc/droneup/duck/daac.cljc)
• uncrew-apollo-frontend repository
• uncrew-missions-service repository
• VT MAAP Chapter 1 Report — DroneUp DAA System Testing (March 2025). Section 4.5.5 Recommendations.
• ASTM F3442M-23 — Standard Specification for Detect and Avoid Performance Requirements