Simulation Distillation: Pretraining World Models in Simulation for Rapid Real-World Adaptation

Jacob Levy, Tyler Westenbroek, Kevin Huang, Fernando Palafox, Patrick Yin, Shayegan Omidshafiei, Dong-Ki Kim, Abhishek Gupta, David Fridovich-Keil

Keywords: Robot Learning: Model Learning, Sim-to-Real Transfer, World Model Adaptation, Robot Manipulation, Quadruped Locomotion

TL;DR: We distill physics simulators into world models from raw perception, then plan and adapt in the real world by finetuning the dynamics

Abstract: Simulation-to-real transfer remains a central challenge in robotics, as mismatches between simulated and real-world dynamics often lead to failures. While reinforcement learning offers a principled mechanism for adaptation, existing sim-to-real finetuning methods struggle with exploration and long-horizon credit assignment in the low-data regimes typical of real-world robotics. We introduce $\texttt{Simulation Distillation}$ ($\texttt{SimDist}$), a sim-to-real framework that distills structural priors from a simulator into a latent world model and enables rapid real-world adaptation via online planning and supervised dynamics finetuning. By transferring reward and value models directly from simulation, $\texttt{SimDist}$ provides dense planning signals from raw perception without requiring value learning during deployment. As a result, real-world adaptation reduces to short-horizon system identification, avoiding long-horizon credit assignment and enabling fast, stable improvement. Across precise manipulation and quadruped locomotion tasks, $\texttt{SimDist}$ substantially outperforms prior methods in data efficiency, stability, and final performance.

Submission Number: 75