Build a Real-Time LiDAR mapper using Edge Impulse and Arduino UNO Q

LDROBOT LiDAR D500 Developer Kit

Arduino® UNO™ Q 4GB

Edge Impulse Studio

<html><body><h3>Classifying Environments with LiDAR and Edge AI on Arduino UNO Q</h3>I'm starting my robotics journey and I wanted to tackle navigation using a scalable approach. Initially I was thinking with the camera however I started with a LiDAR, so I developed this workflow to create useful ML models that can run on Arduino UNO Q, into drones or outdoor robots.This project explores the intersection of LiDAR dToF sensing and Edge AI to classify rooms without the need for compute-intensive computer vision. The methodology lies in transforming 360 degree distance profiles from a 2D dToF LiDAR into time-series feature vectors. Using Edge Impulse Studio, these 'distance fingerprints' are used to train a neural network optimized for the Arduino UNO Q. The result is an efficient machine learning model capable of real-time space recognition at the edge. <figure><img src="https://projects.arduinocontent.cc/e4dc9030-0ddc-40e3-a75d-a49403c422ea.png"/><figcaption></figcaption></figure> Hardware Specifications - Arduino UNO Q - DFRobot D500 LiDAR Developer Kit- USB-Hub Software- <a href="https://github.com/edgeimpulse/example-lidar-mapper-edge-impulse">Python Application for Arduino UNO Q</a>Getting startedThe system architecture follows a structured three-tier approach, starting with the DFRobot D500 LiDAR sensor utilizes dToF technology to perform a continuous 360° scans with a 12-meter range. This data is processed by a Python application running in the Arduino UNO Q capable of both data collection and real-time inference using the Edge Impulse `.eim` engine. Supporting this is the ML Pipeline, where Edge Impulse Studio manages the data ingestion via the Edge Impulse Ingestion API and the feature extraction to neural network training, quantization, and final model deployment into the Arduino UNO Q.<figure><img src="https://projects.arduinocontent.cc/bcf9565d-d388-4c3d-a28e-d5a2a50217ef.png"/><figcaption></figcaption></figure>The application built here processes 360-degree LiDAR scans by capturing 10-second data windows and uploading them via the Edge Impulse Ingestion API. Inside Edge Impulse Studio the datasets are split into 1 second segments, creating 360-dimensional feature vectors where each dimension represents a single degree of the scan.<figure><img src="https://projects.arduinocontent.cc/0c9ea549-a485-4291-b139-86ceb56041c8.png"/><figcaption></figcaption></figure> Once training is finalized, the model is exported as an `.eim` file and copied into the Arduino UNO Q models folder (check the <a href="https://github.com/edgeimpulse/example-lidar-mapper-edge-impulse#copy-the-project-to-the-arduino-uno-q">github repository</a> for more information). After setting the necessary execution permissions, the Python inference engine prepares the model in the edge. As the LiDAR sensor streams new data, the ImpulseRunner executes rapid forward passes to provide real-time classification and show the result in the LiDAR mapper UI.<figure><img src="https://projects.arduinocontent.cc/b8af20da-f170-4c56-80a8-73d24f3043ec.png"/><figcaption></figcaption></figure>My initial tests confirm classification accuracy and low latency in controlled environments. The strength of this approach is its reliance on "distance signatures" distinguishing spaces based on a physical structure rather than the heavy processing requirements of computer vision with cameras and heavy machine learning models. Since this is my first robotics project, I would love to hear your thoughts. Please share your feedback and suggestions in the comments below :point-down: </body></html>

Build a Real-Time LiDAR mapper using Edge Impulse and Arduino UNO Q

Learn how to make an Arduino UNO Q LiDAR Mapper for room recognition