2 Hits in 1.5 sec

LazyDAgger: Reducing Context Switching in Interactive Imitation Learning [article]

Ryan Hoque, Ashwin Balakrishna, Carl Putterman, Michael Luo, Daniel S. Brown, Daniel Seita, Brijen Thananjeyan, Ellen Novoseller, Ken Goldberg
2021 arXiv   pre-print
We present LazyDAgger, which extends the interactive imitation learning (IL) algorithm SafeDAgger to reduce context switches between supervisor and autonomous control.  ...  In physical fabric manipulation experiments with an ABB YuMi robot, LazyDAgger reduces context switches by 60% while achieving a 60% higher success rate than SafeDAgger at execution time.  ...  However, a key challenge in interactive imitation learning is to reduce the burden that interventions place on the human supervisor [24, 53] .  ... 
arXiv:2104.00053v2 fatcat:a4w4gtrh35b6pn774j45ohu22e

ThriftyDAgger: Budget-Aware Novelty and Risk Gating for Interactive Imitation Learning [article]

Ryan Hoque, Ashwin Balakrishna, Ellen Novoseller, Albert Wilcox, Daniel S. Brown, Ken Goldberg
2021 arXiv   pre-print
Effective robot learning often requires online human feedback and interventions that can cost significant human time, giving rise to the central challenge in interactive imitation learning: is it possible  ...  ThriftyDAgger uses a learned switching policy to solicit interventions only at states that are sufficiently (1) novel, where the robot policy has no reference behavior to imitate, or (2) risky, where the  ...  The authors were supported in part by the Scalable Collaborative Human-Robot Learning (SCHooL) Project, NSF National Robotics Initiative Award 1734633, and by donations from Google, Siemens, Amazon Robotics  ... 
arXiv:2109.08273v1 fatcat:uvbfrpd3tbhqxbqvfgkjhvvxn4