Tim Hertweck

About

Tim Hertweck’s research lies at the intersection of reinforcement learning and robotics, with a focus on enabling robots to learn complex motor skills directly from real-world data. His major contributions center on improving data efficiency and generalization through hierarchical and compositional learning. Hertweck pioneered methods like Regularized Hierarchical Policy Optimization (RHPO) and Hindsight Off-policy Options (HO2), which allow robots to transfer learned skills across tasks and learn reusable sub-policies with minimal data. His work on simultaneously learning vision and feature-based control policies for the Ball-in-a-Cup task demonstrated fast, real-world policy training. With over 70 citations across his top papers, Hertweck’s impact is evident in advancing autonomous skill acquisition, from locomotion to stacking diverse objects. Notably, his 2024 paper on mastering stacking through large-scale iterative reinforcement learning on real robots showcases his commitment to practical, scalable solutions. Hertweck’s research is a cornerstone for students and researchers aiming to bridge the gap between sample-efficient algorithms and real-world robotic deployment.

Research Focus

Key Achievements

Top Papers

1. 1
   Simultaneously Learning Vision and Feature-Based Control Policies for Real-World Ball-In-A-Cup

   15 citations · 2019
2. 2
   Regularized Hierarchical Policies for Compositional Transfer in Robotics.

   14 citations · 2019
3. 3
   Towards General and Autonomous Learning of Core Skills: A Case Study in Locomotion

   10 citations · 2020
4. 4
   Data-efficient Hindsight Off-policy Option Learning

   9 citations · 2020
5. 5
   Compositional Transfer in Hierarchical Reinforcement Learning

   7 citations · 2020
6. 6
   Representation Matters: Improving Perception and Exploration for Robotics

   6 citations · 2021
7. 7
   Data-efficient Hindsight Off-policy Option Learning

   6 citations · 2021
8. 8
   Simultaneously Learning Vision and Feature-based Control Policies for Real-world Ball-in-a-Cup

   6 citations · 2019
9. 9
   Simple Sensor Intentions for Exploration

   3 citations · 2020
10. 10
    Mastering Stacking of Diverse Shapes with Large-Scale Iterative Reinforcement Learning on Real Robots

    2 citations · 2024

Key Collaborators

Contact & Links