WHALE: Towards Generalizable and Scalable World Models for Embodied Decision-making

World models play a crucial role in decision-making within embodied environments, enabling cost-free explorations that would otherwise be expensive in the real world. To facilitate effective decision-making, world models must be equipped with strong generalizability to support faithful imagination in out-of-distribution (OOD) regions and provide reliable uncertainty estimation to assess the credibility of the simulated experiences, both of which present significant challenges for prior scalable approaches. This paper introduces WHALE, a framework for learning generalizable world models, consisting of two key techniques: behavior-conditioning and retracing-rollout. Behavior-conditioning addresses the policy distribution shift, one of the primary sources of the world model generalization error, while retracing-rollout enables efficient uncertainty estimation without the necessity of model ensembles. These techniques are universal and can be combined with any neural network architecture for world model learning. Incorporating these two techniques, we present Whale-ST, a scalable spatial-temporal transformer-based world model with enhanced generalizability. We demonstrate the superiority of Whale-ST in simulation tasks by evaluating both value estimation accuracy and video generation fidelity. Additionally, we examine the effectiveness of our uncertainty estimation technique, which enhances model-based policy optimization in fully offline scenarios. Furthermore, we propose Whale-X, a 414M parameter world model trained on 970K trajectories from Open X-Embodiment datasets. We show that Whale-X exhibits promising scalability and strong generalizability in real-world manipulation scenarios using minimal demonstrations.

Any-step Dynamics Model Improves Future Predictions for Online and Offline Reinforcement Learning

Model-based methods in reinforcement learning offer a promising approach to enhance data efficiency by facilitating policy exploration within a dynamics model. However, accurately predicting sequential steps in the dynamics model remains a challenge due to the bootstrapping prediction, which attributes the next state to the prediction of the current state. This leads to accumulated errors during model roll-out. In this paper, we propose the Any-step Dynamics Model (ADM) to mitigate the compounding error by reducing bootstrapping prediction to direct prediction. ADM allows for the use of variable-length plans as inputs for predicting future states without frequent bootstrapping. We design two algorithms, ADMPO-ON and ADMPO-OFF, which apply ADM in online and offline model-based frameworks, respectively. In the online setting, ADMPO-ON demonstrates improved sample efficiency compared to previous state-of-the-art methods. In the offline setting, ADMPO-OFF not only demonstrates superior performance compared to recent state-of-the-art offline approaches but also offers better quantification of model uncertainty using only a single ADM.

Episodic Return Decomposition by Difference of Implicitly Assigned Sub-Trajectory Reward

Real-world decision-making problems are usually accompanied by delayed rewards, which affects the sample efficiency of Reinforcement Learning, especially in the extremely delayed case where the only feedback is the episodic reward obtained at the end of an episode. Episodic return decomposition is a promising way to deal with the episodic-reward setting. Several corresponding algorithms have shown remarkable effectiveness of the learned step-wise proxy rewards from return decomposition. However, these existing methods lack either attribution or representation capacity, leading to inefficient decomposition in the case of long-term episodes. In this paper, we propose a novel episodic return decomposition method called Diaster (Difference of implicitly assigned sub-trajectory reward). Diaster decomposes any episodic reward into credits of two divided sub-trajectories at any cut point, and the step-wise proxy rewards come from differences in expectation. We theoretically and empirically verify that the decomposed proxy reward function can guide the policy to be nearly optimal. Experimental results show that our method outperforms previous state-of-the-art methods in terms of both sample efficiency and performance.

Model-based Reinforcement Learning with Multi-step Plan Value Estimation

A promising way to improve the sample efficiency of reinforcement learning is model-based methods, in which many explorations and evaluations can happen in the learned models to save real-world samples. However, when the learned model has a non-negligible model error, sequential steps in the model are hard to be accurately evaluated, limiting the model's utilization. This paper proposes to alleviate this issue by introducing multi-step plans to replace multi-step actions for model-based RL. We employ the multi-step plan value estimation, which evaluates the expected discounted return after executing a sequence of action plans at a given state, and updates the policy by directly computing the multi-step policy gradient via plan value estimation. The new model-based reinforcement learning algorithm MPPVE (Model-based Planning Policy Learning with Multi-step Plan Value Estimation) shows a better utilization of the learned model and achieves a better sample efficiency than state-of-the-art model-based RL approaches.