You may like to visit Stefan Schaal’s Computational Learning and Motor Control Lab. at the University of Southern California (USC). Beside many papers on the application of RL, supervised ML, nonlinear control, and etc in robotics domain, you can find some nice movies too. For instance, I like the one that robot learns to imitate pole-balancing so much (here!).
I should read a few of those papers soon. I guess I read a paper by this group (Peters and et. al.) about natural policy gradient about a year ago.