notesum.ai

Published at November 19

Reinforcement Learning with Action Sequence for Data-Efficient Robot Learning

cs.AI

cs.RO

Released Date: November 19, 2024

Authors: Younggyo Seo¹, Pieter Abbeel¹

Aff.: ¹UC Berkeley

Arxiv: http://arxiv.org/abs/2411.12155v1

Refer to caption

Task	Length	Demos	Task	Length	Demos
Move Plate	300	51	Cupboards Close All	620	53
Move Two Plates	550	30	Reach Target Single	100	30
Saucepan To Hob	440	28	Reach Target Multi Modal	100	60
Sandwich Flip	620	34	Reach Target Dual	100	50
Sandwich Remove	540	24	Dishwasher Close	375	44
Dishwasher Load Plates	560	17	Wall Cupboard Open	300	44
Dishwasher Load Cups	750	58	Drawers Open All	480	45
Dishwasher Unload Cutlery	620	29	Wall Cupboard Close	300	60
Take Cups	420	32	Dishwasher Open Trays	380	57
Put Cups	425	43	Drawers Close All	200	59
Flip Cup	550	45	Drawer Top Open	200	40
Flip Cutlery	500	43	Drawer Top Close	120	51
Dishwasher Close Trays	320	62