Andrea Baisero

University of Stuttgart, Northeastern University

Papers

Total Citations

H-Index

About

Andrea Baisero is a researcher whose work sits at the intersection of robot learning, human-robot interaction, and reinforcement learning. Her research focuses on making robot programming more intuitive and efficient, particularly through learning from demonstration and feedback, and on tackling the fundamental challenges of reinforcement learning in partially observable environments. Her most cited work, "Robot programming from demonstration, feedback and transfer" (2015, 39 citations), introduces a novel, complementary approach that combines multiple instruction methods to improve the precision and intuitiveness of robot assembly tasks. More recently, Baisero has made significant contributions to reinforcement learning under partial observability. Her 2022 paper, "Leveraging Fully Observable Policies for Learning under Partial Observability," proposes a method to use state information available in simulators to improve learning in real-world, partially observable domains. She has also explored the use of symmetry as an inductive bias for sample-efficient learning in "Equivariant Reinforcement Learning under Partial Observability" (2024). Her work on hierarchical reinforcement learning under mixed observability further demonstrates her commitment to developing robust, theoretically grounded methods for complex robot learning problems.