Agibot Releases Genie Master: Embodied Agent Framework for Continuous Robot Self-Evolution
Agibot has officially released the Genie Master embodied agent framework, which enables robots to continuously learn new skills while working. Powered by this framework, the Spirit G2 robot successfully achieved autonomous self-evolution in real-world retail scenarios without relying on human demonstrations or debugging. This release marks the industry's first verification of Physical RSI (Physical Recursive Self-Evolution) in real business scenarios rather than simulated environments. It represents a major step forward for cross-scenario replication and large-scale commercial deployment of embodied AI robots. Equipped with the Genie Master framework, the Spirit G2 has been validated in convenience stores performing tasks such as cooked food packaging, item picking, and restocking. Agibot also scheduled a 24-hour continuous convenience store operation live test utilizing six Spirit G2 units.
## BACKGROUND
Embodied AI is an interdisciplinary cutting-edge field in artificial intelligence and robotics that emphasizes an intelligent agent interacting dynamically with the physical environment through a body. Unlike pure software AI models, embodied AI systems integrate perception, action, and cognition to autonomously operate in the real world.