The Beijing Humanoid Robot Innovation Center announced today that its open-source dataset, RoboMIND, has achieved a twofold surge in global downloads within just one month, surpassing the 20 million milestone. This places the platform at the top of the industry in both data download volume and market-oriented delivery capability.

Since its open-source release in December 2025, RoboMIND has continuously collaborated with universities, research institutes, and industry partners on multiple technical co-development projects. The initiative has unlocked over 300,000 dual-arm operation trajectories, covering more than 700 practical tasks, with full access granted to developers worldwide. This facilitates the efficient implementation of embodied intelligence perception, motion control algorithm research, and scenario validation for a broad range of R&D teams.

Additionally, within less than a year of its establishment, the Beijing Humanoid Data Base has evolved into one of the most comprehensive professional data collection platforms in China, offering the broadest scenario coverage, the richest robot configurations, and the highest data production capacity and quality. Its fully self-developed, closed-loop data platform toolchain integrates the entire pipeline from data collection, cleaning, quality inspection, and annotation to model training, real-machine evaluation, model deployment, and data feedback. Leveraging internally sourced, high-quality datasets, it conducts industry-leading skill model training, providing robotics companies and industry users with data and training services, including data collection, model training, and performance assessment.

Currently, the data base is advancing toward its goal of "1 million hours of high-quality data," aiming to establish the world's first million-hour high-quality embodied intelligence data platform.