DREAM: Differentiable Real-to-Sim-to-Real Engine for Learning Robotic Manipulation
Haozhe Lou, Mingtong Zhang, Haoran Geng, Hanyang Zhou, Sicheng He, Zhiyuan Gao, Siheng Zhao, Jiageng Mao, Pieter Abbeel, Jitendra Malik, Daniel Seita, Yue Wang
June, 2024Abstract
We introduce DREAM, a differentiable real-to-sim-to-real engine for robotic manipulation that learns from real data, adapts a simulation model, and transfers policies back to hardware through differentiable optimization.
Haozhe Lou
External homepage for publication author link.
Mingtong Zhang
External homepage for publication author link.
Haoran Geng
External homepage for publication author link.
Hanyang Zhou
External homepage for publication author link.
Sicheng He
External homepage for publication author link.
PhD Student@USC
Creating digital doubles with the power of differentiable everything
Siheng Zhao
External homepage for publication author link.
Jiageng Mao
External homepage for publication author link.
Pieter Abbeel
External homepage for publication author link.
Jitendra Malik
External homepage for publication author link.
Daniel Seita
External homepage for publication author link.
Yue Wang
External homepage for publication author link.
