FOURUT is an edge-cloud worker safety monitoring prototype designed to detect early health risks, fall events, fatigue conditions, and unsafe air-quality exposure in construction or industrial working environments. The system combines wearable sensing, local edge decision-making, ESP-NOW environmental communication, AWS IoT cloud ingestion, backend classification, data storage, and supervisor alerting.
This repository is organized as a complete academic engineering artifact. It includes firmware, hardware documentation, cloud configuration, Lambda processing logic, test events, architecture documents, phase summaries, and validation screenshots.
The following materials can be attached when submitting or presenting the project:
| Material | Link |
|---|---|
| Demo video | Google Drive demo video |
| Presentation slides | PPT presentation |
| Round 1 report | docs/reports/round1_concept_design.pdf |
| Round 2 report | docs/reports/round2_prototype.pdf |
Replace the two placeholder links above with the final Google Drive video link and presentation link before submission.
The project addresses the problem of worker safety monitoring in environments where fatigue, falls, oxygen-related abnormalities, and particulate-matter exposure may occur. FOURUT uses an edge-first architecture: urgent safety decisions are made locally on the ESP32 gateway, while the cloud layer is responsible for centralized logging, validation, and notification.
The system monitors four groups of indicators:
| Monitoring Area | Main Data | Purpose |
|---|---|---|
| Biometric condition | Heart rate, SpO2, IBI, HRV indicators | Detect oxygen drop, abnormal heart rate, and fatigue risk |
| Motion condition | Acceleration, gyroscope, SVM, posture change | Detect possible fall events |
| Environmental condition | PM1.0, PM2.5, PM10, AQI | Detect unsafe air-quality exposure |
| Alert state | Edge alarm level, fall flag, risk category | Communicate urgent worker-safety events |
The prototype is not a certified medical or industrial safety device. It is an academic proof-of-concept that demonstrates an integrated sensing, edge-processing, cloud-logging, and alerting workflow.
The complete workflow is:
Worker and environment
↓
MAX30102, MPU6050, APM10 PM sensor
↓
ESP32 edge gateway and ESP-NOW environmental node
↓
Local risk assessment and buzzer alarm
↓
MQTT over TLS through AWS IoT Core
↓
AWS Lambda processing
↓
DynamoDB record, S3 historical log, SNS email alert
The edge device does not wait for the cloud before triggering urgent local alarms. This design reduces response latency and keeps the most safety-critical decision close to the worker.
| Layer | Component | Role |
|---|---|---|
| Wearable edge layer | XIAO ESP32-C3 / ESP32-class gateway | Runs FreeRTOS tasks, reads sensors, performs local risk logic, publishes MQTT payloads |
| Biometric sensing | MAX30102 | Measures heart rate, SpO2, and inter-beat interval data |
| Motion sensing | MPU6050 | Provides acceleration and gyroscope data for fall detection |
| Local alarm | Buzzer and cancel button | Alerts the worker and allows acknowledgement for lower-level alarms |
| Environmental node | ESP32S NodeMCU / ESP32 DevKit + APM10 | Reads particulate-matter data and sends packets by ESP-NOW |
| Cloud ingestion | AWS IoT Core | Receives MQTT telemetry, HRV summaries, and alert events |
| Backend processing | AWS Lambda | Normalizes payloads and classifies worker risk level |
| Storage | DynamoDB and S3 | Stores processed records and historical logs |
| Notification | SNS | Sends email alerts for serious or emergency states |
FOURUT_Worker_Health_Safety_Monitoring/
├── cloud/
│ ├── aws_iot/ # AWS IoT policy, rule, and cloud workflow documentation
│ └── lambda/ # Lambda processor and backend documentation
├── docs/
│ ├── architecture/ # System architecture, communication, and power-management docs
│ ├── figures/ # General project figures
│ └── reports/ # Academic report PDFs
├── firmware/
│ ├── edge_gateway_esp32_two_directions/
│ │ └── FOURUT_firmware_two_directions/
│ │ ├── option_A_single_file/ # Compact demonstration firmware
│ │ └── option_B_modular/ # Modular firmware for maintainability
│ ├── environmental_node_espnow/ # ESP32 + APM10 ESP-NOW node
│ └── micropython_prototype/ # Early MicroPython prototype
├── hardware/
│ ├── hardware_design.md # Hardware design explanation
│ └── schematics/ # Circuit/schematic image
├── phases/ # Phase 1, Phase 2, and Phase 3 summaries
├── results/ # Demo screenshots and validation evidence
├── tests/ # Lambda test events and expected behavior
├── LICENSE
└── README.md
The system uses two communication levels:
| Link | Protocol | Purpose |
|---|---|---|
| Environmental node → edge gateway | ESP-NOW | Sends PM1.0, PM2.5, PM10, AQI, sequence number, and node status |
| Edge gateway → AWS IoT Core | MQTT over TLS | Sends telemetry, HRV summaries, and alert events |
| AWS IoT Core → Lambda | IoT Rule | Triggers cloud-side processing |
| Lambda → DynamoDB/S3/SNS | AWS SDK | Stores records and sends supervisor notifications |
The gateway evaluates multiple risk sources:
| Risk Type | Main Logic |
|---|---|
| SpO2 risk | Warning below 90%, critical below 80% |
| Heart-rate risk | Bradycardia below 50 BPM, tachycardia above 120 BPM |
| Fatigue risk | RMSSD below 20 ms for repeated windows; severe exhaustion when RMSSD is below 15 ms and BPM is above 110 |
| Fall risk | Free-fall phase, impact phase, posture change, and stillness confirmation |
| Environmental risk | AQI warning above 75 and danger above 150 |
Fall detection is intentionally multi-stage to reduce false alarms. A fall is confirmed only when the motion sequence satisfies free-fall, impact, posture-change, and inactivity/stillness conditions within the configured timing window.
The Lambda worker receives telemetry or alert payloads and converts them into a standardized risk state:
| Level | State | Meaning | Notification |
|---|---|---|---|
| 0 | NORMAL |
No abnormal condition detected | No |
| 1 | WARNING |
Early abnormal signal requiring monitoring | Optional / deployment-specific |
| 2 | DANGER |
Serious condition requiring supervisor attention | Yes |
| 3 | EMERGENCY |
Critical condition requiring immediate response | Yes |
Processed events are saved for traceability. Email notifications are used only for serious or critical events in the validation tests.
Path:
firmware/edge_gateway_esp32_two_directions/FOURUT_firmware_two_directions/
This firmware runs the wearable edge gateway. It performs sensor acquisition, fall detection, fatigue estimation, local alarm control, ESP-NOW environmental reception, and MQTT publishing.
Two equivalent implementation directions are included:
| Option | Path | Best Use |
|---|---|---|
| Option A: Single-file firmware | option_A_single_file/fourut_edge_freertos_aws.ino |
Presentation, walkthrough, quick debugging |
| Option B: Modular firmware | option_B_modular/ |
Clean repository submission, extension, maintainability |
Both versions follow the same system behavior. Option A is easier to explain during a live demo, while Option B is better for long-term development because the code is separated into modules such as sensors, fall detection, HRV, alerts, MQTT, ESP-NOW, and FreeRTOS tasks.
Default MQTT topics:
worker/W001/telemetry
worker/W001/hrv
worker/W001/alert
Default edge timing and threshold examples:
| Parameter | Value |
|---|---|
| MPU6050 read interval | 20 ms |
| MAX30102 read interval | 1000 ms |
| Cloud publish interval | 5000 ms |
| Alert cooldown | 30000 ms |
| Free-fall threshold | 4.9 m/s² |
| Impact threshold | 29.4 m/s² |
| Posture-change threshold | 35° |
| Fall latch duration | 30000 ms |
Path:
firmware/environmental_node_espnow/
This firmware runs on an ESP32S NodeMCU or ESP32 DevKit connected to an APM10 particulate-matter sensor. It reads PM1.0, PM2.5, and PM10 values, calculates a prototype PM2.5-based AQI, and sends an EnvPacket to the gateway through ESP-NOW.
Key configuration values are located near the top of:
firmware/environmental_node_espnow/esp32_apm10_espnow_node.ino
Before uploading, verify:
#define ESPNOW_WIFI_CHANNEL 13
#define USE_BROADCAST false
uint8_t GATEWAY_MAC[6] = { ... };
#define PM_RX_PIN 16
#define PM_TX_PIN 17
#define PM_BAUD 1200The ESP-NOW Wi-Fi channel and gateway MAC address must match the gateway Serial Monitor output.
Path:
cloud/aws_iot/
This folder documents the AWS IoT configuration, including the IoT policy, topic structure, and rule-based forwarding workflow. The gateway publishes JSON payloads to AWS IoT Core through MQTT over TLS.
Path:
cloud/lambda/
The Lambda function:
- normalizes incoming payloads;
- extracts worker, biometric, environmental, and alarm fields;
- classifies the condition into
NORMAL,WARNING,DANGER, orEMERGENCY; - stores processed results in DynamoDB and S3;
- sends SNS email notifications for serious or emergency states.
Required Lambda environment variables:
DDB_TABLE_NAME
S3_BUCKET_NAME
SNS_TOPIC_ARN
Open one of the gateway firmware options:
firmware/edge_gateway_esp32_two_directions/FOURUT_firmware_two_directions/option_A_single_file/
or:
firmware/edge_gateway_esp32_two_directions/FOURUT_firmware_two_directions/option_B_modular/
Copy the example secrets file:
secrets.example.h → secrets.h
Then fill in local Wi-Fi and AWS IoT credentials:
#define WIFI_SSID "YOUR_WIFI_SSID"
#define WIFI_PASSWORD "YOUR_WIFI_PASSWORD"
#define AWS_IOT_ENDPOINT "your-endpoint-ats.iot.your-region.amazonaws.com"Also paste the AWS Root CA, device certificate, and private key into secrets.h.
Install the required libraries in Arduino IDE or PlatformIO:
DFRobot_BloodOxygen_S
Adafruit_MPU6050
Adafruit_Sensor
WiFi
esp_now
WiFiClientSecure
PubSubClient
ArduinoJson
Upload either the single-file or modular gateway firmware. Open the Serial Monitor and confirm:
Sensor initialization
Wi-Fi connection
AWS IoT MQTT connection
Gateway MAC address
Wi-Fi channel
Telemetry publishing
Use the printed MAC address and Wi-Fi channel to configure the environmental node.
Open:
firmware/environmental_node_espnow/esp32_apm10_espnow_node.ino
Set the ESP-NOW channel and gateway MAC address. Then upload the firmware to the ESP32 environmental node. The Serial Monitor should show APM10 readings and ESP-NOW send status.
Use the test events in:
tests/lambda_events/
The expected scenarios are:
| Test File | Scenario | Expected State |
|---|---|---|
Test_Normal.json |
Stable worker condition | NORMAL |
Test_Danger.json |
Low SpO2 / high-risk condition | DANGER |
Test_Emergency_Fall.json |
Confirmed fall event | EMERGENCY |
Test_Exhaustion.json |
Severe fatigue condition | EMERGENCY |
Validation screenshots are stored in:
results/demo_screenshots/
They include MQTT reception, Lambda tests, DynamoDB records, S3 logs, IoT Rule configuration, and SNS email notification evidence.
For a complete understanding of the project, read the documentation in this order:
| Order | File | Purpose |
|---|---|---|
| 1 | docs/architecture/system_architecture.md |
Complete system model and layer explanation |
| 2 | docs/architecture/architecture_explanation.md |
Design rationale and edge-cloud decision logic |
| 3 | docs/architecture/communication_workflow.md |
ESP-NOW, MQTT, AWS IoT, Lambda, storage, and alert flow |
| 4 | docs/architecture/power_management.md |
Power domains, state machine, and power-planning strategy |
| 5 | hardware/hardware_design.md |
Hardware design and sensor interface explanation |
| 6 | tests/test_event_descriptions.md |
Lambda validation cases and expected outputs |
| Phase | Focus | Main Outcome |
|---|---|---|
| Phase 1 | Concept design and architecture | Defined worker-safety problem, system layers, communication model, and risk states |
| Phase 2 | Functional prototype and AWS workflow | Built an initial edge-cloud prototype with cloud classification, storage, and alerting |
| Phase 3 | Final integration and environmental extension | Added FreeRTOS gateway structure, ESP-NOW environmental node, PM/AQI data, and full validation evidence |
Detailed phase summaries are available in:
phases/phase1_summary.md
phases/phase2_summary.md
phases/phase3_summary.md
The repository includes evidence for the major prototype functions:
| Evidence | Location |
|---|---|
| MQTT message received by AWS IoT Core | results/demo_screenshots/mqtt_test_client_message.png |
| Lambda normal test success | results/demo_screenshots/lambda_test_normal_success.png |
| Lambda danger email test | results/demo_screenshots/lambda_test_danger_email_sent.png |
| DynamoDB saved item | results/demo_screenshots/dynamodb_saved_item.png |
| S3 saved log | results/demo_screenshots/s3_saved_log.png |
| SNS email alert | results/demo_screenshots/sns_email_alert.png |
| IoT Rule SQL | results/demo_screenshots/iot_rule_sql.png |
These files support the claim that the prototype can transmit data, process it in the cloud, store results, and send supervisor notifications.
Do not commit real credentials or private account information. The following files and values should remain local only:
secrets.h
.env
*.pem
*.key
*.crt
AWS private keys
Wi-Fi password
AWS IoT endpoint if private
SNS topic ARN if sensitive
personal email addresses
Use secrets.example.h and placeholder values for public documentation.
The current repository represents a working academic prototype, but several limitations remain:
| Limitation | Explanation |
|---|---|
| Rule-based classification | The system is interpretable but not adaptive to individual worker profiles |
| Prototype AQI calculation | The AQI method should be aligned with the target deployment standard in future work |
| ESP-NOW encryption not enabled | Local environmental packets are not yet encrypted |
| Limited battery telemetry | Environmental node battery reporting is not fully implemented |
| Multi-worker scaling not generalized | Worker IDs, MQTT topics, and thresholds are currently configured for prototype use |
| Not certified for medical/safety deployment | The system should not be used as a certified medical or industrial safety device |
Recommended next improvements include:
- add configurable worker profiles and adaptive thresholds;
- enable ESP-NOW peer encryption;
- improve battery measurement and runtime estimation;
- add dashboard visualization for supervisors;
- support multiple workers and multiple environmental nodes;
- strengthen payload schema validation before cloud processing;
- evaluate false positives and detection latency using repeated physical tests;
- improve enclosure design for wearable stability and sensor contact quality.
This project is distributed under the license included in LICENSE.
This repository is intended for academic demonstration, project assessment, and future engineering development. It combines embedded firmware, wireless communication, cloud processing, validation evidence, and presentation materials into a single worker health and safety monitoring system.