Project 005
Radar-Based Presence & Positioning System
A radar sensor that tracks where people are in a room, sent over WiFi from an ESP32 to a Raspberry Pi for processing.
Overview
This project set out to track people moving around a room without relying on a camera. Cameras feel invasive for something like this, so the system uses a radar sensor instead, producing coordinates rather than an actual image of anyone in the room.
The setup pairs a radar sensor with an ESP32, which sends the data over WiFi to a Raspberry Pi that handles the processing.
Methodology
Sensor & Data Path
The radar sensor, an HLK-LD2450, can track up to 3 people at once and outputs their X/Y coordinates over UART. An ESP32 board reads that stream and forwards it over UDP to a Raspberry Pi, which handles processing and logging of the incoming position data.
Coordinate Decoding
The sensor's coordinate data isn't formatted the way you'd normally expect negative numbers to work. Instead of standard two's complement, it uses a single bit to flag a value as negative, which meant writing custom decoding logic rather than relying on a standard parsing library.
Engineering Decisions
ESP32-S2 Serial Configuration
The specific ESP32 board used here doesn't have a standard second hardware serial port like the classic ESP32 does, since it relies mostly on native USB. Getting it to reliably read the sensor's UART output took some trial and error with the pin and buffer configuration.
UDP Forwarding
UDP was chosen over TCP for sending the position data across the network, trading guaranteed delivery for lower latency, since occasional dropped packets matter far less here than keeping updates close to real time.
Results
The pipeline runs end to end and streams live position data without dropping out. Working through the coordinate decoding was the main technical hurdle. Right now it's a working single-node system that lays the groundwork for something larger.
Path Forward
The next step is scaling this from a single sensor into a small mesh network, one LD2450 node per room throughout the house. Instead of being limited to whatever a single sensor can see, the goal is to stitch position data together across every room to track roommates as they move from space to space, turning an isolated sensor demo into a whole-house presence system. The user will be able to create a map of their own home, set the sensors and each sensor's angle and position, and then the system will show users moving through the house in real time, tracking useful data from their positions.
Team
Independent Project
Caleb Beswick