K-Scale AGI

K-Scale AGI is an open-source humanoid robot project that aims to build a low-cost, customizable general-purpose robot platform through reinforcement learning and modular hardware design, promoting the democratization of embodied intelligence.

Its core technology is reinforcement-learning-based robot motion control that discards traditional preset actions, letting the robot learn autonomously in simulated environments through reward functions to improve generalization in dynamic environments.

The Stompy robot is targeted at about $8,999 for a complete unit. By modular design and 3D-printed assembly, it aims to keep material costs under $10,000, delivering a consumer-grade price.

Primarily for robot developers, researchers, educational institutions, makers, and tech enthusiasts; it also provides customizable robotic application solutions for small and medium-sized enterprises.

The project addresses data bottlenecks in reinforcement learning through an open hardware platform and community collaboration, encouraging users to generate diverse, large-scale physical interaction data from real-world use.

Yes. The project fully open-sources all software, design files, and the robot operating system, including the ksim simulation tool and APIs, to lower the barrier to entry and promote community collaboration.