EgoForce: Robust Online Egocentric Motion Reconstruction via Diffusion Forcing

2026-05-13

AI Summary

Existing approaches struggle to achieve real-time, robust, and long-term full-body motion reconstruction from sparse and noisy first-person viewpoints. This work proposes the first diffusion-based framework tailored for online egocentric motion reconstruction, leveraging an asymmetric temporal noise scheduling strategy and an incremental denoising mechanism to generate temporally coherent and stable full-body motions under strict causal constraints. The method innovatively integrates diffusion forcing and a noise-robust state imputation strategy, effectively overcoming the trade-off between robustness and efficiency inherent in conventional autoregressive or fixed-window approaches. Experimental results demonstrate that the proposed method significantly outperforms both existing online and offline methods in challenging first-person scenarios, enabling highly robust, long-horizon, and real-time motion reconstruction.

Abstract

With recent advances in embodied agents and AR devices, egocentric observations are readily available as input for real-world interactive online applications. However, egocentric viewpoints can only sporadically observe hands, in addition to the estimated head trajectory. We propose EgoForce, an online framework for reconstructing long-term full-body motion from noisy egocentric input. While existing generative frameworks can robustly handle noisy and sparse measurements, they assume a fixed-length observation window is available and are thus not suitable for real-time applications. Faster inference often relies on autoregressive prediction, sacrificing robustness. In contrast, we adopt a diffusion-based method with a temporally asymmetric noise schedule inspired by Diffusion Forcing. Specifically, our approach models temporally evolving uncertainty and incrementally denoises states as new streaming observations arrive. Combined with a noise-robust imputation strategy, EgoForce progressively generates stable and coherent full-body motion under strict causal constraints. Experiments demonstrate that our online framework outperforms existing online and offline methods, enabling long-horizon, full-body motion reconstruction in challenging egocentric scenarios.

Problem

• egocentric motion reconstruction
• online motion generation
• noisy and sparse observations
• full-body motion
• real-time applications

Innovation

• diffusion forcing
• egocentric motion reconstruction
• online motion generation
• temporally asymmetric denoising
• noise-robust imputation