Bart van Marum
Papers
2
Total Citations
20
H-Index
2
About
Bart van Marum is a researcher advancing the frontier of humanoid robotics, with a primary focus on robust locomotion and reinforcement learning. His most-cited work, "Revisiting Reward Design and Evaluation for Robust Humanoid Standing and Walking" (2024), has already garnered 18 citations, underscoring its immediate impact. In this study, van Marum tackles a critical challenge: enabling humanoid robots to stand and walk while reliably rejecting natural disturbances. He systematically investigates how different reward functions in sim-to-real reinforcement learning (RL) influence controller performance, revealing that careful reward design is essential for achieving robustness in real-world deployment. By dissecting the nuances of reward shaping, van Marum provides a principled framework that moves beyond ad-hoc tuning, offering the robotics community actionable insights for training more resilient locomotion policies. His work bridges the gap between simulation and reality, addressing a key bottleneck in deploying humanoid robots outside controlled labs. For students and researchers, van Marum’s contributions highlight the importance of methodological rigor in RL-based robotics, demonstrating that even subtle choices in reward formulation can determine whether a robot stumbles or strides confidently through unpredictable environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2