Axis Robotics, a robotics startup, has announced a significant achievement in its experiment codenamed 'The Little Prince's Rose.' According to ChainCatcher, the company successfully mobilized 12,000 global users to collect 10,000 high-quality trajectory data entries within five days through its proprietary browser-based platform. This experiment demonstrated that building robotic intelligence can be achieved outside traditional, costly laboratory environments through public participation, leading to a large-scale data explosion.
Leveraging the MetaSim architecture, Axis Robotics developed an end-to-end pipeline from task setup to real machine deployment. The company decoupled complex physical simulations from lightweight web interactions, ensuring the collected data is capable of deep processing. The platform employs cross-simulation replay technology, replaying high-frequency physical data from MuJoCo in Isaac Sim and conducting large-scale domain randomization to enhance data diversity.
Currently, the company has successfully implemented real machine replication on the Franka robotic arm using the data provided by ordinary internet users, validating the feasibility of this data collection model for industrial applications.