Project CATCH: Using Robotics to Correct for Bad Throwing Aim

Russell Wong | 2022-05-04

As the name implies, Project C.A.T.C.H. (Catching Autonomy for Thrown Circular Hazards) consists of a robot arm that will autonomously catch balls thrown at it, even if you have bad aim.

A single Kinect camera is used to perceive the surrounding environment, detect the thrown ball with classical image processing techniques, and extract its position in 3D space. Given several measurements over time, the ball's trajectory is predicted from noisy data points assuming a projectile motion model, and an appropriate goal pose is determined within the robot's workspace. Joint commands are sent to the Franka robot arm to move its end effector to the desired goal pose, which is continuously updated as more detections are made while the ball is in flight, to ultimately catch the ball in the net.

This project was developed by Master of Science in Robotic Systems Development (MRSD) students John Harrington, Mohith Sakthivel, Sabrina Shen, and Russell Wong for the course 16-662 Robot Autonomy taught by Dr. Oliver Kroemer at the Carnegie Mellon University Robotics Institute (CMU RI).

This blog post is still under development. More implementation details and code to come soon! In the meantime, please enjoy this demo video of our robot in action. We hope you enjoy watching this video as much as we enjoyed developing this robot :)


I hope you enjoyed this project! If you'd like to support my work, please consider subscribing to my YouTube channel. Any questions/comments can be directed to my email address. Thanks for reading!