Fleet Anomaly Detection — Edge ⇢ Fog ⇢ Cloud (Simulation using Docker)

This repo simulates a edge-based architecture using Python + Docker:

• Edge (STM32/RPi simulated): 3 containers publishing sensor data to MQTT.
• Fog (Industrial PC): aggregates fleet data, computes simple z-score anomalies, publishes actuator commands.
• Cloud (AWS/Azure/GCP simulated): Streamlit dashboard subscribing to MQTT and visualizing fleet status.
• Broker: Eclipse Mosquitto (MQTT).

Quick Start

# 1) cd into the folder
cd EdgeSim_Anomaly

# 2) Launch the stack
sudo docker-compose up --build

# 3) Open the dashboard
#    http://localhost:8501

# 4) Closing the connections
#     Ctrl+C
sudo docker-compose down

You should see edge1/2/3 printing actuator commands when anomalies are detected, and the Cloud dashboard will show real-time reports.

Topics

• Edge publishes:
  • edge/{edge_id}/sensors → JSON with temperature, humidity, pressure, electrical, vibration
  • edge/{edge_id}/heartbeat
• Fog publishes:
  • fog/actuators/{edge_id} → JSON commands (e.g., { "alarm": true })
  • cloud/reports → JSON status per edge
  • cloud/fleet_snapshot → fleet-level snapshot

Tuning

• fog service env:
  • WINDOW_SIZE (default 50 samples)
  • Z_THRESH (default 2.5)
• edge services env:
  • ANOMALY_RATE (probability of anomaly injection per message)

Extending Toward the Paper

This simulation uses the knowledge gained from the paper A general anomaly detection framework for fleet-based condition monitoring of machines by Kilian Hendrickx, Wannes Meert, Yves Mollet, Johan Gyselinck, Bram Cornelis, Konstantinos Gryllias, Jesse Davis https://doi.org/10.1016/j.ymssp.2019.106585

• Replace fog z-score logic with:
  • DTW/W-DTW warping amount between edges (Block 1)
  • Hierarchical clustering + cophenetic correlation to pick clusters (Block 2)
  • Anomaly score = 1 - (cluster_size / fleet_size) (Block 3)
• Keep MQTT topics the same — only swap fog_node.py internals.

Files

• docker-compose.yml
• mosquitto/mosquitto.conf
• edge/ → Dockerfile, edge_node.py
• fog/ → Dockerfile, fog_node.py
• cloud/ → Dockerfile, cloud_app.py

Contributors

• @Shyam Sundar Raju
• @Dwarakesh V
• @Pooja Shree S
• @Sarigasini M