From Temporal-Evolving to Spatial-Fixing: A Keypoints-Based Learning Paradigm for Visual Robotic Manipulation
Kevin Riou,Kaiwen Dong,Kevin Subrin,Yanjing Sun,Patrick Le Callet,Kevin Riou,Kaiwen Dong,Kevin Subrin,Yanjing Sun,Patrick Le Callet
The current learning pipelines for robotics manipulation infer movement primitives sequentially along the temporal-evolving axis, which can result in an accumulation of prediction errors and subsequently cause the visual observations to fall out of the training distribution. This paper proposes a novel hierarchical behavior cloning approach which tries to dissociate standard behaviour cloning (BC)...
