In learning from demonstrations (LfD) for trajectory planning, end-to-end deep learning (DL) methods offer fast inference and adaptability to complex inputs. However, they are prone to cumulative ...