Adapting World Models with Latent-State Dynamics Residuals
This addresses the challenge of transferring RL agents from simulation to real-world environments, particularly for vision-based tasks, though it appears incremental as it builds on existing residual correction approaches.
The paper tackled the problem of simulation-to-reality discrepancies in reinforcement learning by proposing ReDRAW, a method that adapts world models with latent-state dynamics residuals, resulting in effective modeling of dynamics changes and avoidance of overfitting in low data regimes across vision-based MuJoCo domains and a physical robot task.
Simulation-to-reality reinforcement learning (RL) faces the critical challenge of reconciling discrepancies between simulated and real-world dynamics, which can severely degrade agent performance. A promising approach involves learning corrections to simulator forward dynamics represented as a residual error function, however this operation is impractical with high-dimensional states such as images. To overcome this, we propose ReDRAW, a latent-state autoregressive world model pretrained in simulation and calibrated to target environments through residual corrections of latent-state dynamics rather than of explicit observed states. Using this adapted world model, ReDRAW enables RL agents to be optimized with imagined rollouts under corrected dynamics and then deployed in the real world. In multiple vision-based MuJoCo domains and a physical robot visual lane-following task, ReDRAW effectively models changes to dynamics and avoids overfitting in low data regimes where traditional transfer methods fail.