Yi-Hsiang Chang

National Yang Ming Chiao Tung University

Papers

Total Citations

H-Index

About

Yi-Hsiang Chang is a researcher at the forefront of bridging the gap between simulated and real-world environments for autonomous systems. His work centers on deep reinforcement learning, visual semantic segmentation, and imitation learning, with a particular focus on enabling robots to learn complex control tasks without costly real-world data collection. Chang’s most impactful contribution is his pioneering 2018 study, “Virtual-to-Real: Learning to Control in Visual Semantic Segmentation,” which garnered 69 citations and tackled the critical “reality gap” between synthetic and physical visual data—a key bottleneck in robot learning. By leveraging semantic segmentation to transfer policies from simulation to reality, his work has significantly advanced the feasibility of safe, scalable robot training. Additionally, in “Adversarial Exploration Strategy for Self-Supervised Imitation Learning,” Chang introduced a novel framework that uses adversarial techniques to drive exploration without human demonstrations or extrinsic rewards, pairing a deep reinforcement learning agent with an inverse dynamics model. This innovation has opened new pathways for autonomous skill acquisition in unstructured environments. With a growing citation record and a focus on practical, data-efficient learning, Yi-Hsiang Chang is a rising voice in robotics and AI, shaping how machines learn to navigate and act in the physical world.