Tsinghua University Research Team Reports Study on Humanoid Robots Using Soccer for Cooperative Control

TechXplore reported on Sept. 30, 2026, on research into cooperative control using humanoid robots that play soccer, by Yushi Wang and colleagues at Tsinghua University.

The topic, reported as Tsinghua University research, directly addresses the challenge of training the “see while moving” control required for humanoid robots. If it can reduce the burden of tuning on physical robots, it could affect not only research but also the pace of prototype improvements.

This development appears to show humanoid research shifting from competition in single-function capabilities such as walking and grasping toward integrated control that handles vision, posture control and movement decisions simultaneously. Technically, the soccer task is notable for combining tracking a moving object, loss of visibility, delays and balance maintenance into a single control problem. The source says these complex tasks were learned in a virtual environment, making the extent to which this can reduce trial and error on physical robots a key question. See the original article and published paper for details. We have no related coverage, but in the same field, efforts to bring humanoid robots closer to manufacturing and general-purpose work are advancing alongside efforts to refine control in challenging environments such as sports and competitive tasks. The former focuses on implementation, while the latter builds a foundation for control; soccer tasks such as this one fall into the latter category. Public information does not establish whether such research will lead directly to commercialization soon, but the ability to handle visual recognition and real-time decision-making alongside walking is likely to become a differentiating factor. One possible division of labor is for university research, such as that at Tsinghua University, to develop foundational technologies, while companies integrate them into robot platforms, computing resources and mass-production designs. However, this report does not indicate any connection between specific projects. Three points warrant further attention. First, which control method the research used. Second, whether results in a virtual environment can be replicated on physical robots. Third, how broadly the approach can be generalized to tasks beyond soccer.