🤖 AI Summary
Embodied intelligence requires world models capable of perception, action, and dynamic environmental prediction—yet existing approaches lack a unified framework and suffer from weak physical consistency, error accumulation over long horizons, and trade-offs between real-time performance and modeling fidelity. This paper introduces the first unified world model framework tailored for embodied intelligence, establishing a tri-axial taxonomy grounded in functionality, temporal dynamics, and spatial structure. We systematically integrate multi-source data and physics-aware evaluation metrics; propose novel modeling techniques—including sequential simulation-based reasoning, global discrepancy prediction, and disentangled rendering; and conduct comprehensive quantitative benchmarking across state-of-the-art models. All benchmark datasets, evaluation protocols, and implementation code are publicly released. Our work establishes both theoretical foundations and practical paradigms for world modeling in robotics, autonomous driving, and video understanding.
📝 Abstract
Embodied AI requires agents that perceive, act, and anticipate how actions reshape future world states. World models serve as internal simulators that capture environment dynamics, enabling forward and counterfactual rollouts to support perception, prediction, and decision making. This survey presents a unified framework for world models in embodied AI. Specifically, we formalize the problem setting and learning objectives, and propose a three-axis taxonomy encompassing: (1) Functionality, Decision-Coupled vs. General-Purpose; (2) Temporal Modeling, Sequential Simulation and Inference vs. Global Difference Prediction; (3) Spatial Representation, Global Latent Vector, Token Feature Sequence, Spatial Latent Grid, and Decomposed Rendering Representation. We systematize data resources and metrics across robotics, autonomous driving, and general video settings, covering pixel prediction quality, state-level understanding, and task performance. Furthermore, we offer a quantitative comparison of state-of-the-art models and distill key open challenges, including the scarcity of unified datasets and the need for evaluation metrics that assess physical consistency over pixel fidelity, the trade-off between model performance and the computational efficiency required for real-time control, and the core modeling difficulty of achieving long-horizon temporal consistency while mitigating error accumulation. Finally, we maintain a curated bibliography at https://github.com/Li-Zn-H/AwesomeWorldModels.