arXiv 2510.16732
A Comprehensive Survey on World Models for Embodied AI
By Xinqing Li, Xin He, et al.
Published 2025-10-19
Citation lineage
Review the prior work and downstream research connected to this paper.
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 obj…