Decentralizing Embodied AI Data | Xiangyu Chen (IO AI)
Xiangyu Chen of IO AI presents Humanoids Summit on decentralizing embodied AI data, outlining products for robot teleoperation, sensory data collection, and agent based data workflows.
In this talk
- 0:54
Why robotic data is costly
Chen describes robotic data as high quality but expensive to collect. He says scaling collection requires more workers, environments, processes, and investment, while the resulting data varies in quality.
- 1:58
Data collection products
IO AI offers teleoperation equipment that supports more than 80 robot types, sensory data collection devices, and BodyFlow. Chen says the collection devices can be customized to a customer's requirements.
- 2:54
Teleoperation equipment
Chen describes gloves for full body teleoperation, motion capture, and control of more than 20 types of dexterous hands. The system includes force feedback, and the team tested remote operation between China and the United States across roughly 12,000 kilometers.
- 4:36
BodyFlow agent platform
Chen presents BodyFlow as software available through the cloud or local deployment, with libraries and algorithms that users can combine into agents. For data organization, snapshots, and semantic segmentation, users can ask an agent to apply statistical methods and produce results.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.