Final project: Bluetooth Device Localization
Starting the application. You can see the bluetooth beacons are discovered automatically. Then we give each beacon a position in the room. Once at least three beacons have a position they can start discovering devices.
Early stages of the project. In the video you see the ESP32 beacons (blue circles) estimating the position of several devices. When clicking a device you see extra information like it's position coordinates in the room.
Early stages of the project. Devices with a greater accuracy get rendered on top otherwise you can not interact with them on the map.
Early stages of the project. Development of drawing device circles and implementing trilateration.
A real-time indoor positioning system that locates Bluetooth-enabled devices within a room by combining a network of ESP32 beacons, a Raspberry Pi processing hub, and a live React dashboard — all without any dedicated indoor-positioning hardware.
How It Works
A mesh of ESP32 microcontrollers acts as Bluetooth beacons, continuously scanning for nearby devices and capturing their RSSI (Received Signal Strength Indicator) readings. Those readings are streamed over WebSocket to a Raspberry Pi, which aggregates the data and applies trilateration — the same geometric principle used by GPS — to compute each device’s estimated (x, y) position within the room.
Key Features
- Automatic beacon discovery — new ESP32 nodes are detected and added to the map without manual configuration
- Interactive room setup — drag beacons to their physical positions on a 2-D floor plan before positioning begins
- Live device tracking — positions update continuously as signal readings arrive
- Accuracy-aware rendering — devices with higher positional confidence are drawn on top so they remain clickable
- Per-device detail panel — click any tracked device to inspect its computed coordinates and raw signal data
Technical Highlights
- Python backend on the Raspberry Pi handles RSSI aggregation, trilateration math, and WebSocket broadcasting
- React frontend consumes the WebSocket stream and re-renders the canvas on every position update with no full-page refresh
- Trilateration requires a minimum of three beacons with known positions; the UI enforces this before tracking begins
- Signal noise is mitigated with smoothing applied to rolling RSSI windows before feeding values into the positioning algorithm
This project deepened my understanding of IoT system design, real-time data pipelines, and the practical challenges of translating raw radio signals into reliable spatial coordinates.
Technologies Used
- Python
- React
- Data Science
- Raspberry Pi
- ESP32
- Bluetooth
- WebSocket
- IoT
- Data Visualization
- Data Processing
- Trilateration