Simulation of Right Arm Trajectories of the PR2 Robot Arm using the Guided Policy Search algorithm proposed by Finn. et. al. This simulation runs the catkinized ros indigo version (available at of the gps algorithm).
Here, the robot is told to go to a particular set of end-effector points. Its takes roughly 10 iterations and a cost of about -600 before the robot reaches the desired goal.
This was last Summer's pet project where my colleague, Ben Warwick, and I hacked together node.js, socket.io, Alexa lambda skills, and 3D-printed mounts for the P3-DX robot to make it navigate the AR offices via Alexa voice commands or alternatively via a web-UI.