Visual Control

Package to control MAVs using Visual Odometry Systems and Computer Vision

1. Control Based on Visual Odometry

This package assumes you have a Visual Odometry Package running, such as orb_slam2_ros or svo (those are the already tested ones)

Architecture

• The node control sends the actual velocity commands to the MAV, according to a PID controller stabilizing the drone at the position stablished by the topic/viscon/set_position

• manual_control is a teleop_twist_keyboard based node, that controls MAV via keyboard on a terminal, with options to

  • Append positions to a trajectory
  • Save trajectory to a file - this file will be later used by head node to set positions.
  • Activate and deactivate autonomous control (both head and control nodes)

• head and reads a trajectory from a file and publishes local positions to /viscon/set_position topic

Procedure for Tello

• Run tello driver with roslaunch tello_driver tello_node.launch
• Set camera parameters to worse image quality (frequency is key to visual odometry algorithms) with rosrun dynamic_reconfigure dynparam load /tello/tello [viscon]/config/dump.yaml
• Run manual control with roslaunch viscon manual.launch and take it of pressing '{'
• Run visual odometry algorithm (in this case, orb_slam2_ros) with roslaunch orb_slam2_ros orb_slam2_tello.launch (it needs to be cloned from Skyrats repository)
• Watch it's functioning on rqt GUI with rosrun rqt rqt -d [orb_slam2_ros]/ros/config/rviz_config.rviz
• Save positions on the manual controller with 's'
• Save trajectory with 'f'
• run rosrun viscon head.py
• run rosrun viscon control.py
• Finally, press 'a' on the manual controller to enable autonomous mode!

Have fun!

2. Control Based on Computer Vision

We can also control our MAV based on it's position relative to an object detected by it's camera

Architecture

Procedure for Tello

• Run tello driver with roslaunch tello_driver tello_node.launch

• Set camera parameters to worse image quality (frequency is key to visual odometry algorithms) with rosrun dynamic_reconfigure dynparam load /tello/tello [viscon]/config/dump.yaml

• Run manual control with roslaunch viscon manual.launch and take it of pressing '{'

• Run cv_detection's h_node with rosrun cv_detection h_node

  • This node detects an H in the webcam image and publishes its position in /cv_detection/h_detection as a custom message H_info.msg which contains:
    • detected: boolean that shows if H was detecetd
    • center_x: x coordinate of H's center
    • center_y: y coordinate of H's center
    • area_ratio: ratio between H's area and that of the entire image

• Run control node with rosrun viscon cv_control.py

To use dynamic_reconfigure, rosrun rqt_gui rqt_gui -s reconfigure along with the simple_control.py node

Procedure for MAVROS (using simulation package)

• Run simulate.sh script

  • Check if last line is roslaunch simulation H_world.launch

  rosrun simulation simulate.sh

• Run roslaunch viscon cv_control.launch - depends on cv_detection, it runs

  • cv_detection's h_node
  • viscon's run_h_mission.py
  • viscon's cv_control.py

Result should look something like this