Real-IKEA: Physical Fidelity is the Prerequisite for Robust Manipulation

2026-06-07

AI Summary

This work addresses the poor robustness of robotic manipulation policies due to physical discrepancies between simulation and reality, particularly in articulated object interaction. The authors propose Real-IKEA, a high-fidelity simulation framework that accurately reconstructs the contact geometry and dynamics of IKEA handles and knobs through a six-step physics modeling pipeline. A novel bidirectional surface deviation metric quantifies collision mesh accuracy, while calibrated damping and friction parameters further enhance realism. Reinforcement learning policies trained within this framework demonstrate, for the first time, emergent capabilities such as leveraging mechanical advantage to perform “hooking” and “prying” actions. These policies achieve near-human-level robustness when transferred to real-world tasks, underscoring the critical role of high-fidelity asset modeling in enabling effective sim-to-real transfer.

Abstract

Robotic manipulation robustness often founders on the physics gap between simplified simulations and the resistance-laden real world. In this work, we emphasize that physical realism in articulated interaction is an important ingredient for robust policy learning. We present Real-IKEA, a dataset and simulation framework designed with physical accuracy as a first-class goal. Real-IKEA provides 1,079 articulated asset configurations, derived from 83 authentic IKEA handles and knobs processed through a meticulous six-step physical workflow. For contact-geometry accuracy, we introduce a bidirectional surface-deviation metric to quantify collision meshes. For dynamics realism, we establish resistance-calibrated configurations that vary damping and friction. Crucially, we demonstrate through a Reinforcement Learning (RL) policy that high-fidelity assets enable the discovery of robust "hooking" and "levering" strategies that prioritize mechanical advantage over fragile friction-pulling. Together, these results position Real-IKEA as a critical benchmark for developing manipulation policies capable of human-level robustness in articulated object tasks.

Problem

• robotic manipulation
• physics gap
• physical fidelity
• articulated objects
• simulation-to-reality

Innovation

• physical fidelity
• articulated objects
• contact geometry
• dynamics realism
• robust manipulation