Max Robotics — Global Service Robot Directory + US Certification

About

Banafsheh Rafiee is a researcher focused on advancing reinforcement learning (RL) through improved representation learning and real-world robotic applications. Her work addresses a critical bottleneck in deep RL: the tendency of neural networks to over-generalize, which can degrade performance. In her most cited work (2020, 8 citations), she proposes novel methods to "break" harmful generalization, enabling agents to learn more robust and efficient policies. This contribution is significant because it tackles a fundamental challenge in scaling RL beyond toy domains to complex, real-world tasks. Rafiee also explores the intersection of prediction and intelligence, empirically comparing off-policy learning algorithms—including General Value Functions (GVFs)—on physical robots (2019, 4 citations). Her research bridges theory and practice, demonstrating how continual prediction can underpin intelligent behavior. By combining algorithmic innovation with rigorous robotic validation, Rafiee is helping to build more capable, generalizable RL systems. Her work is particularly relevant for researchers in deep RL, robotics, and representation learning, offering practical insights for improving agent performance in dynamic environments.