Robot Follow Human using YOLOv8 & ROS2

A ROS2 package that enables a robot to detect and follow a human in real-time using a camera and YOLOv8 object detection.

Developed for autonomous robots running ROS2 Humble (tested on Ubuntu 22.04, Python 3.10).

---

Overview

This project implements a human-following robot system using YOLOv8 for real-time person detection and ROS2 for robot control and communication. The system processes camera images to detect humans, estimates their position, and publishes movement commands to make the robot follow the detected person autonomously.

It consists of two main ROS2 nodes:

human_position_publisher — runs YOLOv8 to detect humans and publishes their bounding box coordinates.

follow_human — subscribes to detected positions and controls the robot to follow the target at a safe distance.

Optional visualization using RViz2 allows monitoring camera input, detection results, and robot motion in real-time.

ROS2 Packages

Make sure you have:

sudo apt install ros-humble-cv-bridge ros-humble-rclpy ros-humble-sensor-msgs
sudo apt install ros-humble-geometry-msgs ros-humble-launch ros-humble-rviz2

Install Environment

Install ROS dependencies:

sudo apt install ros-humble-cv-bridge ros-humble-image-transport python3-colcon-common-extensions

Install the necessary Python libraries:

cd robot_detect
pip install -r src/requirements.txt

Usage

1. Git clone directory

git clone https://github.com/is-buiquocdoanh/robot_detect.git

2. Build

cd robot_detect
colcon build
source install/setup.bash

2. Launch gazebo simulator and robot model (terminal 1)

cd robot_detect
source install/setup.bash
ros2 launch robot_simulation simulation.launch.py

After running, gazebo will be displayed with pre-set environment, along with differential robot model.

3. Launch models yolov8 (terminal 2)

cd robot_detect
source install/setup.bash
ros2 launch robot_recognition launch_yolov8.launch.py

• Start the YOLOv8 detector node yolov8_ros2_pt.py
• Get image from topic /camera/image_raw
• Detect people and output bounding box coordinates to topic /inference_result
• Display detection log in terminal (position and confidence)

4. Launch follow human

cd robot_detect
source install/setup.bash
ros2 launch follow_human follow_human.launch.py

This command will launch the entire “Follow Human” system, including:

• Node human_position_publisher: detects and outputs the position of the person from the camera.
• Node follow_human: controls the robot to move according to the detected person.
• (Optional) Open RViz2 to visually observe the camera and the direction of the robot's movement.

ROS2 Topics

Topic Name	Message Type	Description
/camera/image_raw	sensor_msgs/Image	Input image stream from the camera
/human_positions	yolov8_msgs/HumanPositionArray	Detected human positions from YOLOv8
/cmd_vel	geometry_msgs/Twist	Velocity command output for robot movement

Control Parameters (in follow_human.py)

Parameter	Description	Default Value
target_area	Target bounding box area — smaller value means the robot keeps a farther distance	0.15
center_x	Image center position (for a 640px wide image)	320.0
linear_k	Linear velocity gain — controls forward/backward movement speed	0.3
angular_k	Angular velocity gain — controls rotation speed toward the target	0.3
stop_threshold	Minimum movement threshold — smaller value reduces oscillation near the target	0.03

---

Author

• Name: BUI QUOC DOANH
• Email: doanh762003@gmail.com
• Github: is-buiquocdoanh

License

This project is released under the MIT License