arxivcs.RO2026-07-22
Learning Diverse Humanoid Tasks via Synthetic Video Scenarios without Real World Data
Yun-Hao Tsai, Cong-Thanh Vu, Yen-Chen Liu
The human-like morphology of humanoid robots grants them exceptional potential for agile and versatile motor capabilities, but it also introduces significant challenges in acquiring complex skills. Traditional Learning-from-Demonstrations methods are often constrained by the high…