Dorian Goepp

Centre National de la Recherche Scientifique, Centre Inria de l'Université de Lorraine

Papers

Total Citations

H-Index

About

Dorian Goepp is a robotics and machine learning researcher whose work centers on the intersection of reinforcement learning and autonomous robotic systems. His most significant contribution lies in advancing data-efficient policy search methods for robotics, a critical challenge in the field where real-world robot interactions are costly, time-consuming, and potentially damaging to hardware. Goepp's most notable work, "Black-Box Data-Efficient Policy Search for Robotics" (2017), addresses one of the fundamental bottlenecks in applying reinforcement learning to physical robotic systems. His research leverages uncertain dynamical models — techniques that allow robots to learn effective control policies from remarkably few real-world episodes by first constructing probabilistic models of robot dynamics and subsequently optimizing policies against those models. This approach dramatically reduces the number of physical interactions required before a robot can perform competently, making practical deployment far more feasible. The paper has accumulated 93 citations, reflecting meaningful recognition within the robotics and machine learning communities. For students entering the field of model-based reinforcement learning or robot learning, Goepp's work represents an important contribution to making intelligent robotic systems more accessible and deployable in real-world environments where data collection remains a persistent constraint.