RealAppliance: Let High-fidelity Appliance Assets Controllable and Workable as Aligned Real Manuals
🤖 AI Summary
Existing household appliance simulation assets suffer from low rendering fidelity, incomplete modeling of physical and electronic mechanisms, and misalignment with official user manuals—resulting in a substantial reality gap between simulation and real-world operation. To address this, we introduce RealAppliance: the first high-fidelity appliance dataset rigorously aligned with official manufacturer documentation, covering 100 appliance categories and comprehensively modeling外观 (visual appearance), physical dynamics, electronic circuits, and program logic. Building upon it, we construct RealAppliance-Bench—a multimodal, multi-task evaluation benchmark supporting manual comprehension, component localization, and open-loop/closed-loop operational planning for embodied AI. Extensive experiments validate its effectiveness in evaluating both multimodal large language models and embodied planning systems. RealAppliance establishes a standardized, realistic testbed and performance benchmark for appliance interaction research.
Application Category
📝 Abstract
Existing appliance assets suffer from poor rendering, incomplete mechanisms, and misalignment with manuals, leading to simulation-reality gaps that hinder appliance manipulation development. In this work, we introduce the RealAppliance dataset, comprising 100 high-fidelity appliances with complete physical, electronic mechanisms, and program logic aligned with their manuals. Based on these assets, we propose the RealAppliance-Bench benchmark, which evaluates multimodal large language models and embodied manipulation planning models across key tasks in appliance manipulation planning: manual page retrieval, appliance part grounding, open-loop manipulation planning, and closed-loop planning adjustment. Our analysis of model performances on RealAppliance-Bench provides insights for advancing appliance manipulation research
Problem
Research questions and friction points this paper is trying to address.
Addresses poor rendering and incomplete mechanisms in appliance assets
Aligns appliance mechanisms with manuals to reduce simulation-reality gaps
Evaluates models on manual retrieval and manipulation planning tasks
Innovation
Methods, ideas, or system contributions that make the work stand out.
High-fidelity appliance dataset with complete mechanisms
Benchmark for evaluating multimodal and embodied models
Tasks include manual retrieval and manipulation planning
🔎 Similar Papers
No similar papers found.
