Reborn Network is creating an open platform for AGI robots by enabling individuals to contribute data and models. The platform aims to turn human behavior into digital assets that can be used to train and improve robots. This allows for a collaborative approach to developing physical intelligence.
Funding
Funding not disclosed
Founders
Product
Problem
Training general-purpose humanoid robots requires vast amounts of diverse and high-quality data, but current datasets are limited in scale and scope, hindering the development of robust embodied AI models. Existing AI models often lack generalization capabilities due to insufficient training signals, making them narrowly applicable and ineffective across different tasks and environments. The diversity in robot hardware designs further complicates the deployment of universal intelligence, as each platform requires specific tuning and adaptation.
Solution
Reborn Network is building a decentralized protocol and open ecosystem designed to address the data, model, and embodiment gaps in robotics. The platform enables individuals to contribute human motion data from various sources, including VR/AR gaming, motion capture using Rebocap™ hardware, and real-world task videos. This collective intelligence is then transformed into valuable training signals for Robotic Foundation Models (RFMs) that can generalize across different robot embodiments and tasks. Reborn's approach facilitates the creation of versatile physical AI models, accessible through the platform, and rewards contributors with tokens, fostering a community-owned approach to AGI robot development.
Target Audience
The primary audience includes robotics companies, AI researchers, and developers focused on training and deploying embodied AI models for humanoid robots, as well as individuals interested in contributing to the development of AGI.
Features
- Unified Data Platform: Aggregates multimodal data streams from embodied vlogs, mocap life, VR gaming, and a simulation engine.
- Rebocap™: Low-cost wearable devices for high-fidelity motion capture.
- Roboverse Simulation: High-fidelity simulation engine for generating synthetic training data.
- Open Model Ecosystem: Growing library of Versatile Physical AI Models, including OpenVLA models, full-body control policies, and dexterous manipulation agents.
- Community-Driven Validation: Data curation and validation through community consensus.
- Token-Based Rewards: Contributors earn Reborn tokens for their participation in the ecosystem.