Nox: A DIY SLAM‑Enabled House‑Wandering Robot Powered by ROS

Nox is a differential‑drive robot built on ROS, Raspberry Pi, and Arduino, equipped with a Kinect 360 for real‑time SLAM and obstacle detection.

Initially conceived as a modular base, I expanded it into a fully autonomous platform capable of creating a map of its surroundings (using gmapping), localizing itself, and planning collision‑free paths to any target location.

Powered by an 11.1 V Li‑Ion battery and two 12 V DC motors (107 rpm) sourced from Banggood, the robot can be re‑armed quickly— the front panel is removable and accepts batteries of various lengths. A battery‑level display alerts you when a recharge is needed.

Although the motors provide ample speed, I chose to prioritize precise encoders for smoother navigation.

The chassis was designed to integrate the Kinect seamlessly. I opted for a sleek, triangular form inspired by contemporary design and the iconic “Deus Ex” aesthetic. Extensive CAD tweaking ensured a professional finish while keeping wiring manageable.

Status LEDs—recycled from a New Year’s Eve glow stick—provide visual cues: a rapid three‑blink sequence signals that the Arduino is not yet connected to the ROS master; a breathing pattern indicates active navigation, with blink rate scaling to robot speed.

Not connected blinking

Idle blinking

Structure

The differential drive places both motors on a shared axis. A wooden base supports two caster wheels for stability; the rear section hosts stacked plates for the Raspberry Pi, Arduino, and other electronics.

The outer enclosure is a hand‑cut black plastic plate, a design that will be streamlined with 3D printing in future builds.

Hardware

At the core sits a Raspberry Pi 3B running Ubuntu and ROS, reachable via Wi‑Fi and SSH for remote control. ROS handles odometry, navigation planning, and SLAM with Kinect data. The Pi sends velocity commands to an Arduino Mega 2560, which drives the motors through an Adafruit Motor Shield using a PID loop. Encoder feedback is fed back to the Pi for accurate odometry.

I experimented with an Uno and Leonardo before settling on the Mega 2560, which offers ample memory and interrupt pins for encoder reads.

The Kinect 360 is central to the project: a 25 € 3D camera that delivers depth data for gmapping, and also provides vision, an accelerometer, and a microphone array—valuable for future expansions.

Power management uses two DC converters: the motors run directly off the 11.1 V battery, the Pi, Arduino, and encoders operate at 5 V, while the Kinect receives a stabilized 12 V.

Software

The project runs ROS Kinetic. The sole custom node, nox_controller, calculates odometry from motor speed data. The kinematic model follows the approach detailed in “On Differential Drive Robot Odometry with Application to Path Planning.” Covariance matrices are published for future sensor fusion.

Motor control currently uses direct PWM commands; the PID loop is disabled but will be calibrated for improved precision. Encoder feedback already yields satisfactory speed regulation.

Read More Detail: Nox – A House Wandering Robot (ROS)