Published in International Conference on Learning Representation, 2024
This paper analyzes the existing data augmentation methoods in Deep Reinforcement Learning and makes recommendations on how to apply data augmentation in DRL.
Recommended citation: Jianshu Hu, Yunpeng Jiang, Paul Weng. "Revisiting Data Augmentation in Deep Reinforcement Learning." ICLR 2024. https://openreview.net/forum?id=EGQBpkIEuu
Published in IEEE Transaction on Robotics, 2024
In this paper, we propose a model optimization algorithm that automatically synthesizes reduced-order models, optimal with respect to a user-specified distribution of tasks and corresponding cost functions.
Recommended citation: Yu-ming Chen, Jianshu Hu, Michael Posa. "Beyond Inverted Pendulums: Task-optimal Simple Models of Legged Locomotio." T-RO 2024. https://arxiv.org/abs/2301.02075
Published in International Conference of Robot Learning, 2022
This paper leverages a learned dynamics model for generating new data to improve the sample efficiency of DRL.
Recommended citation: Jianshu Hu, Paul Weng. "Solving Complex Manipulation Tasks with Model-Assisted Model-Free Reinforcement Learning." CoRL 2022. https://openreview.net/pdf?id=7CrXRhmzVVR