HSImul3R: Physics-in-the-Loop Reconstruction of Simulation-Ready Human-Scene Interactions
2026-03-16 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionRobotics
AI summaryⓘ
The authors introduce HSImul3R, a system that creates accurate 3D models of people interacting with objects from regular photos or videos. Unlike previous methods, their system makes sure these models follow real-world physics, so simulations run smoothly without glitches. They achieve this by using a special process that improves both the person's movements and the objects' shapes by learning from a physics simulator. They also created a new test set called HSIBench to evaluate different scenarios. Their work shows that HSImul3R can make stable reconstructions ready for use in real robots.
3D reconstructionhuman-scene interactionphysics simulationreinforcement learningscene geometrymotion fidelitycontact stabilityphysical constraintssimulation feedbackembodied AI
Authors
Yukang Cao, Haozhe Xie, Fangzhou Hong, Long Zhuo, Zhaoxi Chen, Liang Pan, Ziwei Liu
Abstract
We present HSImul3R, a unified framework for simulation-ready 3D reconstruction of human-scene interactions (HSI) from casual captures, including sparse-view images and monocular videos. Existing methods suffer from a perception-simulation gap: visually plausible reconstructions often violate physical constraints, leading to instability in physics engines and failure in embodied AI applications. To bridge this gap, we introduce a physically-grounded bi-directional optimization pipeline that treats the physics simulator as an active supervisor to jointly refine human dynamics and scene geometry. In the forward direction, we employ Scene-targeted Reinforcement Learning to optimize human motion under dual supervision of motion fidelity and contact stability. In the reverse direction, we propose Direct Simulation Reward Optimization, which leverages simulation feedback on gravitational stability and interaction success to refine scene geometry. We further present HSIBench, a new benchmark with diverse objects and interaction scenarios. Extensive experiments demonstrate that HSImul3R produces the first stable, simulation-ready HSI reconstructions and can be directly deployed to real-world humanoid robots.