Diffusion World Model: Future Modeling Beyond Step-by-Step Rollout for Offline Reinforcement Learning
This addresses the challenge of efficient future modeling for offline RL practitioners, offering a novel approach to reduce recursive queries and improve value estimation.
The paper tackles the problem of long-horizon prediction in offline reinforcement learning by introducing the Diffusion World Model (DWM), which predicts multistep future states and rewards in a single forward pass, resulting in a 44% performance gain over one-step dynamics models and competitive results with model-free methods.
We introduce Diffusion World Model (DWM), a conditional diffusion model capable of predicting multistep future states and rewards concurrently. As opposed to traditional one-step dynamics models, DWM offers long-horizon predictions in a single forward pass, eliminating the need for recursive queries. We integrate DWM into model-based value estimation, where the short-term return is simulated by future trajectories sampled from DWM. In the context of offline reinforcement learning, DWM can be viewed as a conservative value regularization through generative modeling. Alternatively, it can be seen as a data source that enables offline Q-learning with synthetic data. Our experiments on the D4RL dataset confirm the robustness of DWM to long-horizon simulation. In terms of absolute performance, DWM significantly surpasses one-step dynamics models with a $44\%$ performance gain, and is comparable to or slightly surpassing their model-free counterparts.