mecanum_ros2_yolo

This represents the final project for my thesis. It features 4WD Mecanum wheels controlled via Raspberry Pi Pico (micro-ROS) with custom kinematics/odometry. ROS2 Humble on Raspberry Pi 4B integrates RPLiDAR A1 (SLAM via slam_toolbox) and Pi Camera Module 3 with custom YOLOv8 detection. Includes from-scratch implementations of A path planning and Pure Pursuit path tracking for full autonomy.

🙏 Acknowledgements

This project builds upon the excellent work of Jon Durrant and his “DDD Diff Drive Robot” series for Raspberry Pi Pico:

• GitHub Repository
• YouTube Playlist

I adapted and extended Jon’s micro-ROS motor-control code to support a 4-wheel Mecanum configuration, redesigned the kinematics/odometry, and integrated a full ROS 2 navigation stack for omnidirectional autonomy.

🎥 Demo

Real-Time Inference–Based Navigation

Obstacle‑Aware Path Planning & Tracking

🔧 Hardware

• Chassis & Drive
  • 4× Mecanum wheels
  • 4x DC Motors
  • 2x Motor driver boards (TB6612FNG)
• Boards
  • Raspberry Pi Pico (motor control via micro-ROS)
  • Raspberry Pi 4 B (ROS 2 Humble)
• Sensors
  • RPLiDAR A1
  • Raspberry Pi Camera Module 3

📦 Directory Structure

• PID_Testing
  • firmware/ Contains the standalone Raspberry Pi Pico firmware for PID loop validation.
• ROS_Implementation
  • firmware/ Micro-ROS agent and motor‐control code for the Pi Pico.
  • ros2_ws/
    • droid1/
      • config/ - YAML files for the joystick controller.
      • launch/ - Python launch scripts for SLAM, sensing, and navigation.
      • rviz/ - RViz configuration for visualizing robot state & map.
      • urdf/ - URDF/XACRO robot description.
    • tf_pub/ Custom ROS 2 package that computes and publishes the transform tree between odom and base_link, based on your Pico’s odometry output.
    • yolobot_recognition/ Autonomous perception & navigation stack: YOLOv8‐based object detector node (Pi Camera integration) and A\* path planner & Pure Pursuit controller launch scripts

🚀 Deployment

Prerequisites

This project is built and tested on Ubuntu 22.04 LTS with ROS2 Humble LTS.

Setup and build the worskpace

cd $HOME
git clone https://github.com/cristianooo1/mecanum_ros2_yolo.git
cd ~/mecanum_ros2_yolo/ros2_ws
colcon build --symlink-install
source install/setup.bash

Launch the microros agent on the Raspberry Pi 4B

sudo micro-ros-agent serial --dev /dev/ttyACM0 baudrate=115200

Launch the LiDAR module on the Raspberry Pi 4B

ros2 launch rplidar_ros rplidar_a1_launch.py

Launch the camera module on the Raspberry Pi 4B

ros2 run v4l2_camera v4l2_camera_node --ros-args -p image_size:="[640,480]" -p camera_frame_id:=camera_link_optical

Launch the slam_toolbox module on the Raspberry Pi 4B

ros2 launch droid1 online_async_launch.py slam_params_file:=src/droid1/config/mapper_params_online_async.yaml

Launch RVIZ

cd /src/droid1
ros2 launch droid1 dispOnly.launch.py model:=urdf/AM.urdf.xacro

Launch the TF publisher

ros2 run tf_pub tf_pub

Launch the real time YOLO inference script

ros2 run yolobot_recognition yolov8_ros2_pt.py

Launch the script to calculate the 3D position of the detected object

ros2 run yolobot_recognition detect_object_3D.py

Launch the script to follow the detected object

ros2 run yolobot_recognition yolov8_follower.py

Launch the script to calculate the trajectory to a target position

ros2 run yolobot_recognition path_planning.py

Launch the script to follow the calculated trajectory

ros2 run yolobot_recognition pure_pursuit_omni.py