Autonomous Wind Turbine Inspection Framework Enabled by Visual Tracking Nonlinear Model Predictive Control (VT-NMPC)

This repository contains the code and simulation files for the paper titled "Visual Tracking Nonlinear Model Predictive Control Method for Autonomous Wind Turbine Inspection". Link to paper

For this purpose, a time optimal path planner and a Visual tracking MPC is developed.

We provide a general inspection framework, that takes the dimensions of the wind turbine as input, and provides the optimal attitude rate and thrust command to the drone to acheive optimal coverage.

The approach is modular, where the global plan for inspecting is provided through a time optimal graph based path planner. The output of the path planner is sequentially input to a NMPC with visual tracking costs, that allows the drone to acheive optimal pose relative to the surface for best heading, incidence angle and distance from the surface. More details on the method can be found soon through the paper submitted for publication.

Installation instructions:

1. Install Ubuntu 18.04 and ROS Melodic.

2. Clone directory in the home folder:

   cd
   git clone [email protected]:open-airlab/VTNMPC-Autonomous-Wind-Turbine-Inspection.git

3. Download the PX4 folder from here and place it inside the Wind-Turbine-Inspection folder.

4. Copy the folder WTI_catkin inside the catkin_ws.

5. Download MAVROS dependencies:

   sudo apt-get install ros-melodic-mavros*
   sudo apt-get install xdotool
   sudo apt-get install ros-mavros-mav-msgs

6. Build workspace:

   cd catkin_ws
   catkin_make

7. Setup PX4:

   cd Wind-Turbine-Inspection
   ./install_dependencies_and_setup_px4_modified.sh

   Note: Ignore the errors related to Python 2.7.

8. Add aliases for arming the drone and setting the mode to offboard:

   sudo gedit ~/.bashrc

   Add the following lines:

   alias arm='rosrun mavros mavsafety arm'
   alias disarm='rosrun mavros mavsafety disarm'
   alias offboard='rosrun mavros mavsys mode -c OFFBOARD'

Starting the simulation:

cd Wind-Turbine-Inspection/WTI_px4_modified/shell_scripts/
./run_sitl_gazebo_withWrapper_terminator.sh matrice_100

Running the Inspection Planner:

1. Launch RQT Reconfigure:

   roslaunch dji_m100_trajectory m100_trajectory_v2_indoor.launch

2. Activate traj_on.

3. In the terminal, arm the drone and set the mode to offboard:

   arm
   offboard

   The drone will take off.

Launching the VT-NMPC:

roslaunch quaternion_point_traj_nmpc quaternion_point_traj_nmpc.launch

Adding wind to the simulation:

roslaunch dji_m100_trajectory windgen_recdata.launch

Running the whole inspection framework:

The optimal sequence of points and surface normals are created in txt file format, which the NMPC uses to generate optimal control actions. The planner is run for a default wind turbine model.

To run the default planner:

1. Bring the drone to the initial position (-3, 0, 3).

2. Run the point and normal generator node:

   rosrun dji_m100_trajectory GP_statemachine

3. Change the mode to GP (Global Planner) and tick "point to inspect" checkbox.

Citation

If you use this framework in your work, please cite the following paper: Link to paper

@inproceedings{amer2023visual,
  title={Visual Tracking Nonlinear Model Predictive Control Method for Autonomous Wind Turbine Inspection},
  author={Amer, Abdelhakim and Mehndiratta, Mohit and le Fevre Sejersen, Jonas and Pham, Huy Xuan and Kayacan, Erdal},
  booktitle={2023 21st International Conference on Advanced Robotics (ICAR)},
  pages={431--438},
  year={2023},
  organization={IEEE}
}

Star History