Software and Control Architecture of a Ball Catching Drone

This repository hosts a ROS workspace of the control and software architecture of a ball catching drone in a Gazebo simulation. All details can be found in ME4232_Final_Report.pdf with a summarised version in our poster, ME4232poster.pdf. The catkin workspace was made in ROS Melodic (Ubuntu 18.04).

We attempt to intercept the path of a moving ball

with a drone in simulation

via a depth camera

After filtering the image using colour thresholding and fringe point removal by removing points that are over 2 standard deviations from the center of the detected ball, we get the following result in simulation:

We then consider multiple possible path planning options for the drone. The first test being a simple detection and collision with a stationary ball

The next method is to simply chase after the ball in a ‘cat chase mouse’ method where the drone moves towards a moving ball and yaws if it starts to leave the view of the depth camera.

The issue with this method is that it is not very efficient, so a better way would be to predict the path of the ball via trajectory prediction and then plan the path of the drone as such.

One example would be to intercept the trajectory of the ball at the point closest to the drone.

But the issue with this is that with time, the actual path will deviate from the predicted path.

Hence it may be better to intercept earlier along a ball’s path even if it may not be the closest to the drone.

The dashboard of the control & software architecture in the Ubuntu 18.04 terminal is as shown

The RQT graph of all ROS topics and nodes is as shown:

Collaborators

Arijit Dasgupta

Jasper Tan

Arjun Agrawal