TurtleBot-RL-Navigation

Train a TurtleBot3 to navigate using reinforcement learning and PyTorch

Overview

This project trains a TurtleBot3 robot to navigate through a maze using DQN agent. It uses laser scan data and outputs movement commands to reach its goal.

Key Technologies

Ros Noetic Gazebo Sim TurtleBot3 Burger PyTorch Custom Maze World

Installation

Prerequisites

Ubuntu 20,04 ROS Noetic Gazebo 11 Python 3.8+

Install ROS packages

sudo apt update sudo apt install-ros-noetic-turtlebot3* ros-noetic-gazebo-ros-pkgs ros-noetic-xacro

Clone and build the workspace

cd ~/catkin_ws/src git clone https://github.com/adnansanaulla/TurtleBot-RL-Navigation.git turtlebot_rl cd ~/catkin_ws catkin_make source devel/setup.bash

Set TurtleBot3 model

echo "export TURTLEBOT3_MODEL=burger" >> ~/.bashrc source ~/.bashrc

Python dependencies

pip3 install -r ~/catkin_ws/src/turtlebot_rl/requirements.txt

Training the agent

roscore // in a new terminal: roslaunch turtlebot_rl train_world.launch // in another terminal: python3 ~/catkin_ws/src/turtlebot_rl/scripts/train.py

Evaluation

python3 ~/catkin_ws/src/turtlebot_rl/scripts/eval.py