Shiyu Wu

Papers

Total Citations

105

H-Index

About

Shiyu Wu is a robotics researcher whose work sits at the intersection of robot learning, manipulation, and perception. His most influential contribution, "Learning Behavior Trees From Demonstration" (2019), has garnered 78 citations and represents a significant advance in Learning from Demonstration (LfD) — a paradigm that empowers non-experts to program robots for complex, multistep tasks. By applying behavior trees as a structured learning framework, Wu's approach moves beyond low-level primitive actions to enable more generalizable and interpretable task representations, broadening the accessibility of robot programming. Wu has also made notable strides in robotic perception through his work on transparent object manipulation. His "GlassLoc" system introduces plenoptic imaging to detect grasp poses among transparent objects in cluttered environments — a notoriously difficult open challenge in robotics, accumulating over 27 citations across publications. By leveraging light-field camera technology, GlassLoc addresses the deep uncertainty that transparent materials introduce for robot vision systems. Together, Wu's research reflects a coherent vision: making robots more capable of operating reliably in real-world environments, whether by learning from human demonstration or by perceiving challenging physical materials with greater accuracy.