X-Humanoid Unveils New Robotics Platform
Beijing's Innovation Center of Humanoid Robotics (X-Humanoid) has launched its latest general-purpose robot platform, the Embodied Tien Kung 3.0. The new system is designed with a focus on enhanced openness and practical usability for developers and researchers in the humanoid robotics field.
- The Beijing Innovation Center of Humanoid Robotics, also known as X-Humanoid, was established in November 2023 in Beijing's Economic-Technological Development Area to accelerate the industrialization of humanoid robot technology. - X-Humanoid recently secured a significant strategic investment of 700 million RMB (approximately £77 million) from investors including Baidu, which will be used for further development and large-scale promotion of its core technologies. - The Embodied Tien Kung 3.0 stands 169 cm tall, weighs 62 kg, and is equipped with 43 degrees of freedom. Its processing power comes from an Intel Core i7 processor and dual Nvidia Jetson AGX Orin boards, delivering around 550 trillion operations per second. - A key feature of the Tien Kung 3.0 is its open ecosystem; the hardware includes multiple expansion interfaces, and the software supports common protocols like ROS2, MQTT, and TCP/IP, allowing for easier secondary development by partners. - This new version demonstrates significant advancements in dynamic movement, capable of climbing over 1-meter obstacles, performing complex actions in confined spaces, and even executing a one-handed vault over a block. - Predecessor models have been tested in real-world scenarios, including a version that ran a 100-meter sprint in 21.50 seconds and another that was deployed on an unmanned production line at a Foton Cummins engine plant. - The robot operates on the "Huisi Kaiwu" general embodied intelligence platform, which creates a "perception-decision-execution" closed loop for more autonomous operations rather than relying on pre-programmed routines. - As part of its open approach, X-Humanoid has open-sourced several key technologies, including the robot's physical design blueprints, a vision-language model (Pelican-VL), and the RoboMIND dataset to help combat market fragmentation.