The Beijing Innovation Center of Humanoid Robotics (X-Humanoid) has had eight original research papers accepted by the 10th Conference on Robot Learning (CoRL 2026), covering key areas of humanoid robotics, including data generation and augmentation, whole-body motion, all-terrain locomotion, whole-body manipulation, online learning, real-time control and operational safety.
The papers collectively address eight technical challenges involved in moving embodied AI from research into real-world applications. According to the center, this marks the largest group of papers accepted from a Chinese humanoid robotics team in the history of the conference.
CoRL is a leading international academic conference in robot learning. Its double-blind peer-review process and selective acceptance standards attract research teams from universities, laboratories, and industry worldwide, making it an important forum for advances in embodied AI and robotic learning.
The eight papers reflect the center's approach to developing integrated systems rather than isolated capabilities. While some robotics research focuses primarily on locomotion and other work on manipulation, the center aims to connect these capabilities with data generation, control, continuous learning, and safety constraints.
Built around its "HuiSi KaiWu" platform, which integrates large and small AI models for humanoid robots, X-Humanoid has developed a technical framework spanning data generation and augmentation, whole-body movement and all-terrain mobility, hierarchical control and execution, online learning and iteration, and safety measures for real-world deployment. The center says it is among a small number of domestic humanoid robotics teams with capabilities covering the full stack of embodied AI, from theory and algorithms to validation on physical robots.
In 2026, more than 20 of the center's original research papers have been accepted by leading international conferences in robotics and artificial intelligence, including the International Conference on Learning Representations (ICLR), the IEEE International Conference on Robotics and Automation (ICRA), the International Conference on Machine Learning (ICML), the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), and CoRL.
The research spans embodied foundation models, robotic manipulation and control, motion planning and safety constraints. The center plans to make key findings from the eight newly accepted papers available as open-source resources through the Huisi Kaiwu platform. The research will also be used to improve the performance of its Tiangong series humanoid robots in competitions and industrial applications.