Roberto Rama

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

Papers

Total Citations

H-Index

About

Roberto Rama is a robotics and machine learning researcher whose work sits at the intersection of reinforcement learning and autonomous systems. His most influential contribution centers on data-efficient policy search for robotics, a critical challenge in the field where real-world robot learning is constrained by the high cost and time required to gather training data. Rama's research advances a class of algorithms that leverage uncertain dynamical models to dramatically reduce the number of episodes needed to train effective robot policies. Rather than relying on brute-force data collection, his approach iteratively learns a model of robot dynamics after each episode and uses optimization techniques to derive policies that maximize expected performance — enabling robots to learn meaningful behaviors with far fewer interactions with the physical world. This work, published in 2017, has accumulated over 90 citations, reflecting its meaningful uptake within the robotics and machine learning communities. The research addresses a fundamental bottleneck in deploying learning-based systems on physical hardware, making it particularly relevant for applications where data collection is expensive or dangerous. Rama's contributions represent an important step toward making reinforcement learning practically viable for real-world robotic systems.