Maria Huegle

Papers

Total Citations

H-Index

About

Maria Huegle is a researcher advancing the frontiers of reinforcement learning, with a core focus on improving data efficiency and robustness in off-policy methods for real-world robot control. Her major contributions target the fundamental limitations of Deep Q-learning, which often struggles with sample inefficiency and sensitivity to environmental noise. In her highly cited work, "Off-policy Multi-step Q-learning" (2019), she introduced a multi-step TD-learning approach to enhance data efficiency, a critical step for deploying RL in physical systems. Expanding on this, her paper "Composite Q-learning: Multi-scale Q-function Decomposition and Separable Optimization" (2019) proposes a novel framework that decomposes the Q-function into multiple scales, enabling separable optimization to mitigate the adverse effects of stochasticity in rewards and environments. Though early in her career, these works have garnered over 7 citations, signaling their growing influence. Huegle’s research is particularly notable for its direct applicability to robotics, where efficient and stable learning is paramount. Her innovative decomposition strategy offers a promising pathway toward more reliable and practical reinforcement learning systems, marking her as a rising voice in the field.