Tsu-Ching Hsiao

Papers

Total Citations

H-Index

About

Tsu-Ching Hsiao is a robotics and machine learning researcher whose work sits at the intersection of computer vision, sim-to-real transfer, and robotic manipulation. Their most recognized contribution, "Virtual-to-Real: Learning to Control in Visual Semantic Segmentation" (2018), addresses one of the most persistent challenges in robot learning: bridging the reality gap between simulated training environments and real-world deployment. By leveraging visual semantic segmentation, Hsiao's approach enables robots to transfer policies learned in simulation to physical settings without costly or hazardous real-world data collection — a contribution that has garnered over 75 citations and influenced subsequent work in domain adaptation for robotics. Building on this foundation, Hsiao's more recent research pushes the boundaries of precision robotics through diffusion-based methods. Their 2024 paper, "Precise Pick-and-Place using Score-Based Diffusion Networks," introduces a coarse-to-fine continuous pose diffusion framework that significantly enhances object pose estimation accuracy for robotic manipulation tasks. Together, these works reflect a sustained commitment to making robot learning safer, more efficient, and more practically deployable — contributions of growing relevance as autonomous systems become increasingly embedded in real-world environments.