OASIS: From Simulation Data Collection to Real-World Humanoid Loco-Manipulation

2026-06-07

AI Summary

This work addresses the longstanding challenge in humanoid robotics where high-quality real-world trajectory data is scarce yet essential for scalable mobile manipulation. The authors propose OASIS, a framework that leverages 3D generative models to reconstruct object assets from a single real image, enabling realistic simulation environments. Within this simulated setting, teleoperated trajectories are collected and augmented through domain randomization to enhance data diversity, which in turn trains a hierarchical visuomotor policy. Remarkably, OASIS achieves zero-shot transfer to real humanoid robots using only simulated data, outperforming policies trained on real teleoperation data across most tasks. This breakthrough demonstrates that generative simulation plays a pivotal role in bridging the sim-to-real gap and significantly enhances policy robustness without requiring real-world fine-tuning.

Abstract

Recent progress in robot manipulation has been largely driven by learning from large-scale demonstrations. For humanoid robot loco-manipulation tasks, however, existing data sources force an unsatisfying tradeoff between trajectory quality and scalability. Real-world teleoperation provides the highest-quality trajectories but requires dedicated physical space and time-consuming scene resets. Simulation offers an alternative way out of this dilemma: it can produce clean, embodiment-aligned data at scale without any physical hardware. In this paper, we propose OASIS, a simulation-data-driven framework for humanoid loco-manipulation. OASIS automatically reconstructs realistic object assets from real-world images using a 3D generative model. Based on these assets, trajectories are first collected through teleoperation in simulation, and then augmented under diverse domain randomizations in a post-processing stage. With the resulting simulation data, we further design a hierarchical visuomotor policy for humanoid loco-manipulation. Extensive experiments on the real humanoid robot show that, under zero-shot deployment, the policy trained on our simulation data achieves higher success rates on most tasks than that trained on real-robot teleoperation data, owing largely to the broad lighting and environmental variations covered by our simulation rendering, which real-robot data fails to capture. The project page is available at https://oasis-humanoid.github.io/.

Problem

• humanoid loco-manipulation
• simulation data
• teleoperation
• data scalability
• trajectory quality

Innovation

• simulation-to-reality
• 3D generative modeling
• domain randomization
• humanoid loco-manipulation
• visuomotor policy