Mingdong Wu

I am a PhD student in the School of Computer Science in Peking University, advised by Prof. Hao Dong. I received my bachelor degree in 2021, from Turing Class in Peking University.

My research interest lies in robot learning, generative models, and reinforcement learning. I aim to build intelligent embodied agents that automatically discover and perform tasks without human specification.

Email  /  Google Scholar


TarGF: Learning Target Gradient Field for Object Rearrangement
Mingdong Wu, Fangwei Zhong, Yulong Xia, Hao Dong.
NeurIPS, 2022
paper link / project page / bibtex

We study object rearrangement without explicit goal specification. The agent is given examples from a target distribution and aims at rearranging objects to increase the likelihood of the distribution. Our key idea is to learn a target gradient field that indicates the fastest direction to increase the likelihood from examples via score-matching. We further incoporates the target gradient field with reinforcement learning or model-based planner to tackle this task in model-free and model-based setting respectively. Our method significantly outperforms the state-of-the-art methods in the quality of the terminal state, the efficiency of the control process, and scalability.