X-Humanoid Open-Sources Tien Kung 3.0 Robot
X-Humanoid has launched and is open-sourcing its Embodied Tien Kung 3.0 humanoid robotics platform. The company stated it will make the robot's body design, controls, models, and datasets publicly available. The move is intended to democratize access to advanced robotics and build a collaborative development ecosystem.
- The robot's onboard computing consists of dual NVIDIA Jetson AGX Orin boards, providing approximately 550 TOPS for processing visual and tactile data, alongside a high-performance Intel Core i7 processor for logical reasoning and system management. - It runs on the "Wise KaiWu" embodied AI platform, which uses a continuous perception-decision-execution loop; it employs Vision-Language Models (VLMs) for scene interpretation and planning, and Vision-Language-Action (VLA) models for real-time autonomous navigation and control. - The open-source release is extensive, including the RoboMIND dataset, the ArtVIP simulation asset library, training toolchains, and the underlying world models and VLM/VLA models. The software stack supports common communication middleware and protocols like ROS2, MQTT, and TCP/IP to simplify integration. - Physically, the robot features high-torque integrated joints that enable millimeter-level operational accuracy and dynamic movements like clearing one-meter obstacles. It also has multiple expansion interfaces to support the flexible integration of various end-effectors and tools for different industrial tasks. - The Tien Kung platform has a track record of advanced mobility; a previous version demonstrated its capabilities by climbing 134 uneven outdoor stairs in a public park using vision-based walking, a significant step beyond controlled lab environments. Another model, the Tien Kung Ultra, won the first-ever humanoid robot half-marathon. - X-Humanoid, also known as the Beijing Humanoid Robot Innovation Center, is a state-backed Chinese research institute established in November 2023 as a joint venture by companies including Xiaomi Robotics and UBTech.
