PULSE Collector — React Native App

Data Collection Layer for PULSE 2.0 AMD Slingshot Hackathon — Smartphone Sensor Pipeline

---

Architecture

📱 PULSE Collector (React Native / Expo)
│
├── 5 Sensor Channels (simultaneous)
│   ├── Accelerometer + Gyroscope @ 200Hz  →  IMU packets
│   ├── GPS @ 1Hz                           →  GPS packets  
│   ├── Camera (rear) @ 2fps               →  FRAME packets (base64 JPEG)
│   └── Microphone @ 10Hz (RMS)            →  AUDIO packets
│
├── Transport
│   ├── WebSocket → ws://<laptop>:8000/ws/<session_id>
│   └── SQLite offline buffer (auto-drain on reconnect)
│
└── On-device IRI
    └── Rolling RMS estimate for live display (full computation on backend)

---

Directory Structure

pulse-collector/
├── App.jsx                      ← Navigation root
├── app.json                     ← Expo config + permissions
├── package.json
│
├── screens/
│   ├── SetupScreen.jsx          ← Server config, session name, camera height
│   ├── RecordingScreen.jsx      ← MAIN: live sensor streams + record button
│   └── HistoryScreen.jsx        ← Past sessions, export, delete
│
├── hooks/
│   ├── useIMU.js                ← Accelerometer + Gyroscope @ 200Hz
│   ├── useGPS.js                ← Location watcher + distance tracking
│   ├── useCamera.js             ← Rear camera frame capture @ 2fps
│   └── useAudio.js              ← Microphone RMS via expo-av metering
│
├── services/
│   ├── WebSocketClient.js       ← WS connect/reconnect/queue singleton
│   └── OfflineBuffer.js         ← SQLite sessions + packet + segment tables
│
├── components/
│   ├── SensorStatusBar.js       ← 5-dot sensor status indicators
│   ├── AccelWaveform.js         ← SVG sparkline of accel Z
│   ├── IRIGauge.js              ← Large IRI number with condition badge
│   └── SegmentHistory.js        ← Horizontal chips of completed 100m segments
│
└── utils/
    ├── theme.js                 ← Colors, spacing, IRI color helpers
    └── iriEstimate.js           ← Rolling RMS IRI + haversine distance

---

Setup

1. Install dependencies

cd pulse-collector
npm install

2. Start Expo

npx expo start

Scan QR code with Expo Go app (iOS/Android).

For production build:

npx expo run:ios
npx expo run:android

3. Backend must be running

On your laptop:

cd pulse/backend
uvicorn main:app --host 0.0.0.0 --port 8000

Both devices must be on the same WiFi network.

---

WebSocket Packet Format

All packets are JSON objects sent over ws://<host>/ws/<session_id>.

IMU Packet (200Hz)

{
  "type": "IMU",
  "timestamp": 1709123456789,
  "ax": 0.12,
  "ay": -0.03,
  "az": 9.83,
  "rx": 0.01,
  "ry": 0.00,
  "rz": 0.02
}

ax/ay/az in m/s² · rx/ry/rz in rad/s

GPS Packet (~1Hz)

{
  "type": "GPS",
  "timestamp": 1709123456790,
  "lat": 12.34567,
  "lng": 78.90123,
  "speed_ms": 8.33,
  "speed_kmh": 30.0,
  "accuracy_m": 4.2,
  "heading": 270.0,
  "distance_m": 450.2
}

Frame Packet (2fps)

{
  "type": "FRAME",
  "timestamp": 1709123456800,
  "data": "/9j/4AAQSkZJRgAB...",
  "width": 640,
  "height": 480
}

data is base64-encoded JPEG

Audio Packet (10Hz)

{
  "type": "AUDIO",
  "timestamp": 1709123456850,
  "rms": 0.042,
  "dbfs": -27.5,
  "sample_rate": 44100
}

Backend → App (Segment Complete)

{
  "type": "SEGMENT_COMPLETE",
  "segment": {
    "segment_id": "12.3456,78.9012",
    "iri_value": 4.8,
    "iri_condition": "Poor",
    "rut_depth_mm": 23.1,
    "surface_type": "WBM",
    "final_condition": "Poor",
    "gps": { "lat": 12.3456, "lng": 78.9012 }
  }
}

---

Physical Setup (Critical for IRI accuracy)

1. Mount phone rigidly — windshield suction mount or dashboard holder
2. Rear camera facing road — unobstructed view of road surface
3. Measure camera height — from phone lens to road surface (typically 1.1–1.4m in a car)
4. Enter camera height in Setup screen before starting
5. Drive ≥ 20 km/h for valid IRI readings

---

IRI Classification (IRC:SP:20)

IRI (m/km)	Condition	Action
< 2.0	Good	Routine Maintenance
2.0 – 4.0	Fair	Preventive Treatment
4.0 – 6.0	Poor	Rehabilitation
> 6.0	Very Poor	Reconstruction

---

Permissions Required

Permission	Platform	Purpose
Motion/Accelerometer	iOS (NSMotionUsageDescription)	IRI computation at 200Hz
Location	iOS + Android	GPS speed + segment geotag
Camera	iOS + Android	Visual road surface assessment
Microphone	iOS + Android	Acoustic surface classification
Wake Lock	Android	Keep screen on during recording

---

Offline Mode

When the backend is unreachable:

• IMU/GPS/Audio packets are queued in memory (up to 500 packets)
• Session metadata and segment results are always saved to SQLite
• On reconnection, queued packets are drained to backend automatically
• Session history is always accessible offline

---

PULSE 2.0 — AMD Slingshot Hackathon, February 2026