RESEARCH PAPER

RoboScape: Physics-informed Embodied World Model

Yu Shang; Xin Zhang; Yinzhou Tang; Lei Jin; Chen Gao; Wei Wu; Yong Li

Classification

View four quadrants
Architecture
Not applicable
Prediction paradigm
Not applicable
Source review status
Not assigned

Category review. Action-conditioned RGB/depth prediction supplies a learned visual environment for external policy rollouts and synthetic policy-training data. Actions are provided by another policy, rather than predicted by RoboScape itself. The cataloged contribution is a dataset/data-generation method or evaluation/simulation resource, not the architecture of an evaluated or external policy. Reading evidence

AT A GLANCE

Contribution

RoboScape predicts action-conditioned robotic RGB and depth videos using coupled autoregressive branches. Depth-feature feedback and motion-selected keypoint supervision aim to improve geometry and interaction dynamics. Reported benefits include stronger video metrics, useful synthetic policy-training data, and correlated simulator-based policy evaluation; the ablations expose tradeoffs, and physical robot deployment remains future work.

Abstract

An abstract has not been added yet.

Affiliations

Tsinghua University; Manifold AI