The most difficult part of deploying home service robots is the evaluation stage; a high-fidelity simulation system combined with real-world scene data makes for a powerful combination.

Mars Finance News: Recently, the top international computer vision conference CVPR 2026 announced its Highlight research achievements. A research team from Peking University, in collaboration with Shangwei Qiyuan Research Institute and JD.com, officially launched the RealAppliance dataset and the accompanying RealAppliance-Bench evaluation benchmark, focusing on the challenge of smart appliance operation planning driven by instruction manuals. They use a high-fidelity simulation system to address the practical evaluation issues of household service robots, promoting embodied intelligence to accelerate its deployment in real home environments.