PR2 Robotic Nerf Sentry running ROS This was a senior project for a class in robot learning where we trained a PR2 classify between threatening and non-threatening targets via their physical pose, trained with a SVM. Then we use a Kalman filter to predict the naive linear motion of the human target to account for the relatively slow nerf bullets.