/Mobile-Robot-Navigation-Using-Deep-Reinforcement-Learning-and-ROS

Deep Reinforcement Learning Based Mobile Robot Navigation Using ROS2 and Gazebo

Primary LanguagePythonMIT LicenseMIT

DRL-for-Mobile-Robot-Navigation-Using-ROS2

Table of Contents

  1. Project Structure
  2. Requirements
  3. Build
  4. Training
  5. Testing
  6. Additional Demos
  7. Contribution

Project Structure

.
├── 📂 docs/: contains demo videos
│   ├── 📄 dynamic_environment.mp4
│   ├── 📄 slam.mp4
│   └── 📄 simulation.mp4
├── 📂 drl_agent/: main deep reinforcement learning agent directory
│   ├── 📂 config/: contains configuration files
│   ├── 📂 launch/: contains launch files
│   ├── 📂 scripts/: contains code for environment, policies, and utilities
│   └── 📂 temp/: stores models, logs, and results
├── 📂 drl_agent_description/: contains robot description files, models, and URDFs
│   ├── 📂 launch/: launch files for agent description
│   ├── 📂 meshes/: 3D models of the robot
│   ├── 📂 models/: contains specific model files for kinect sensors
│   └── 📂 urdf/: URDF files for camera, laser, and robot description
├── 📂 drl_agent_gazebo/: contains Gazebo simulation configuration and world files
│   ├── 📂 config/: simulation and SLAM configuration files
│   ├── 📂 launch/: Gazebo launch files for various setups
│   ├── 📂 models/: Gazebo models used in the simulation
│   └── 📂 worlds/: simulation worlds for training and testing environments
├── 📂 drl_agent_interfaces/: custom action, message, and service definitions
│   ├── 📂 action/: defines DRL session actions
│   ├── 📂 msg/: empty for now
│   └── 📂 srv/: service definitions for environment and robot interactions
├── 📂 velodyne_simulator/: Velodyne LiDAR simulation setup

Requirements

Other requirements

pip install -r requirements.txt

Build

  • Clone this repository: 
    mkdir -p ~/drl_agent_ws/src
cd ~/drl_agent_ws/src
git clone --recurse-submodules git@github.com:anurye/DRL-for-Mobile-Robot-Navigation-Using-ROS2.git .
  • Install dependencies: 
    cd ~/drl_agent_ws
rosdep install --from-path src -yi --rosdistro humble
  • Build the workspace: 
    cd ~/drl_agent_ws
colcon build

Training

  • Export the environment variable DRL_AGENT_PACKAGE_PATH:

    echo 'export DRL_AGENT_PACKAGE_PATH=~/drl_agent_ws/src/drl_agent' >> ~/.bashrc
source ~/.bashrc

  • Launch the simulation:

    Terminal 1:

    cd ~/drl_agent_ws
source install/setup.bash
ros2 launch drl_agent_gazebo simulation.launch.py

    Terminal 2:

    cd ~/drl_agent_ws
source install/setup.bash
ros2 run drl_agent_gazebo environment.py

    Terminal 3:

    cd ~/drl_agent_ws
source install/setup.bash
ros2 run drl_agent train_td7_agent.py

Testing

If you have closed the terminals, restart the simulation in Terminal 1 and Terminal 2 as described above.

Terminal 3:

cd ~/drl_agent_ws
source install/setup.bash
ros2 run drl_agent test_td7_agent.py

Additional Demos

Contribution

Feel free to contribute or ask questions by opening an issue