README.md

ROS2 Reinforcement Learning Drive

This repository provides a customizable ROS2 environment for training multiple 2D drive cars using Deep Q-Network (DQN). Currently, agents learn collision-free navigation by avoiding obstacles, but the environment is designed to allow easy modification to train for various objectives in the future.

Occupancy Map launch

ros2 launch reinforcement_learning_drive occupancy.launch.py

Gazebo Map launch

ros2 launch reinforcement_learning_drive gazebo.launch.py

DQN model run

ros2 run reinforcement_learning_drive_model dqn_node

Demo

• Single Actor Youtube Video
• Multi Actor Youtube Video