Found 1 document(s)
Page 1 of 1
-
Vision-based insertion requires correcting alignment errors while retaining useful behavior learned from demonstrations. We present ACT--Flow, an online learning framework that couples an adaptive Action Chunking Transfo...