Search papers, labs, and topics across Lattice.
This paper introduces EgoGenesis, an egocentric world-action simulator that synthesizes high-quality manipulation videos to enhance the training data for embodied AI. By employing Online Anchored Projective Memory and Action-3D Rotary Position Embedding, EgoGenesis achieves improved visual fidelity and action alignment in long egocentric rollouts. The method significantly boosts real-robot success rates, demonstrating a notable increase in generalization for both single-arm and dual-arm tasks when augmented with synthesized trajectories.
Synthesized egocentric videos can elevate real-robot task success rates by over 10% through enhanced training data.
Egocentric video offers rich manipulation experience for embodied AI, yet collecting diverse egocentric data across scenes, objects, motions, and embodiments remains costly. We present \method, an egocentric world-action simulator that synthesizes controllable, high-quality manipulation videos to expand scarce real-world training data. \method{} builds on a pretrained video generation prior and introduces two geometry-aware conditioning mechanisms. Online Anchored Projective Memory (OAPM) preserves a first-frame 3D scene anchor while periodically refreshing a recent state during autoregressive generation. Action-3D Rotary Position Embedding (A3D-RoPE) encodes end-effector motion with camera-aware 3D rotary coordinates, injecting action geometry into skeleton-to-video cross-attention for precise control. Together, these components improve visual fidelity, geometric stability, and action alignment in long egocentric rollouts. Moreover, augmenting 400 real trajectories with 400 \method-generated trajectories improves out-of-distribution real-robot success from 77\% to 84\% on single-arm tasks and from 53\% to 70\% on dual-arm tasks, demonstrating that the synthesized data substantially improve downstream WAM generalization.
DAMO