Max Robotics — Global Service Robot Directory + US Certification

About

Eric Tzeng is a researcher whose work sits at the intersection of robot learning, domain adaptation, and computer vision. His research focuses on enabling robotic systems to transfer learned representations across different environments — a critical challenge when real-world data collection is costly or dangerous. One of his most influential contributions explores adapting deep visuomotor representations from simulated to real-world settings, recognizing that simulation offers a scalable but imperfect proxy for physical deployment. This line of work, which has accumulated over 60 citations, laid important groundwork for sim-to-real transfer in robotic control. His most cited contribution (81 citations) extended this framework using weak pairwise constraints, offering a more flexible approach to domain adaptation. Earlier in his career, Tzeng also contributed to learning from demonstrations for deformable object manipulation, tackling the complex problem of scene registration and trajectory optimization — work that reflects his broader interest in generalizable robot learning. Collectively, his research addresses one of robotics' most persistent bottlenecks: making learned policies robust enough to function reliably when conditions inevitably shift from training to deployment.