On August 20, during the 2026 World Robotics Conference, Optical Wheel Intelligence released the first 100,000-hour full-modal human behavior open source data set EgoSuite-Open100K, the RoboFinals pilot site simulation evaluation infrastructure, and the RoboStack real scene deployment and feedback platform. Together, the three form a self-contained intelligent continuous learning system. Among them, EgoSuite-Open100K covers 128 scenarios and more than 15,000 tasks, and the first batch of data will be simultaneously open sourced with Hugging Face and AtomGit; RoboFinals provides large-scale concurrent assessments for industrial manufacturing, logistics and warehousing, power inspection, special operations, commercial services, and medical care to make up for shortcomings in pilot testing; and RoboStack focuses on the actual work process after deployment to support automatic collection and return of field failure samples and rapid model iteration.

On August 20, during the 2026 World Robotics Conference, Optical Wheel Intelligence released the first 100,000-hour full-modal human behavior open source data set EgoSuite-Open100K, the RoboFinals pilot site simulation evaluation infrastructure, and the RoboStack real scene deployment and feedback platform. Together, the three form a self-contained intelligent continuous learning system. Among them, EgoSuite-Open100K covers 128 scenarios and more than 15,000 tasks, and the first batch of data will be simultaneously open sourced with Hugging Face and AtomGit; RoboFinals provides large-scale concurrent assessments for industrial manufacturing, logistics and warehousing, power inspection, special operations, commercial services, and medical care to make up for shortcomings in pilot testing; and RoboStack focuses on the actual work process after deployment to support automatic collection and return of field failure samples and rapid model iteration.