Penalty Save

Each Section contains the ground truth demonstration an adaptive prior should adapt to, the averaged behaviors or the meta-trained prior as the three main method in used in the paper, and the generated roll-outs when truly imitating the ground truth demonstration from scratch.

demo_vid1.mp4

meta_prior_env1.mp4

meta_prior.mp4

vid1_lr1_nokl.mp4

demo_vid3.mp4

meta_prior_env3.mp4

vid1_lr1_nokl.mp4

meta_prior.mp4

vid3_lr1_nokl.mp4

meta_prior.mp4

vid3_lr1_nokl.mp4

demo_vid5.mp4

meta_prior_env1.mp4

vid5_lr1_nokl.mp4

meta_prior.mp4

vid5_lr1_nokl.mp4

meta_prior.mp4

vid5_lr1_nokl.mp4

demo_vid4.mp4

vid5_lr1_nokl.mp4

vid4_lr1_nokl.mp4

meta_prior.mp4

vid4_lr1_nokl.mp4

meta_prior.mp4

vid4_lr1_nokl.mp4