Jan Metz

About

Jan Metz is a leading researcher in robot learning, with a primary focus on solving the challenge of reward function specification for autonomous systems. His major contributions lie in the field of active reward learning, where he develops methods that enable robots to efficiently infer human preferences through targeted queries. His seminal 2014 paper, "Active Reward Learning" (88 citations), established foundational techniques for addressing the "reward design problem" in tasks like robotic grasping, where reliable success metrics are often unavailable. Metz advanced this work with his 2015 study, "Active reward learning with a novel acquisition function" (50 citations), which introduced an innovative query selection strategy to maximize learning efficiency. Together, these highly cited papers have shaped how researchers approach human-robot interaction and preference-based policy optimization. By reducing the need for hand-crafted reward functions, Metz's work has made robot learning more practical and scalable, directly impacting applications in manipulation, autonomous navigation, and assistive robotics. His research continues to bridge the gap between theoretical reinforcement learning and real-world robotic deployment.

Research Focus

Key Achievements

Top Papers

1. 1
   Active Reward Learning

   88 citations · 2014
2. 2
   Active reward learning with a novel acquisition function

   50 citations · 2015

Key Collaborators

Contact & Links