World’s First "Force-Position Hybrid Control Algorithm" Proposed

Recently, a Chinese research team achieved a major breakthrough in the field of robotic algorithms by proposing the world’s first "unified theory of force-position hybrid control algorithms." This algorithm enables robots to simultaneously learn position and force control without relying on force sensors, increasing the success rate of related tasks by approximately 39.5% compared to strategies using only position control. What’s more noteworthy is that the related paper has already won the Outstanding Paper Award at the International Conference on Robot Learning, marking the first time the award has been won by an all-Chinese team of scholars since its establishment.

At the experimental site of the Beijing Institute for General Artificial Intelligence, a quadruped robotic dog equipped with this new algorithm was methodically performing a training task of wiping a whiteboard. Using this practical scenario, researchers explained the core principles and outstanding advantages of the force-position hybrid control algorithm to reporters.

Researchers explained that the widely used Vision-Language-Action (VLA) models often struggle with many real-world tasks, primarily because these tasks mostly involve complex contact scenarios. For example, when wiping a blackboard, a robotic arm must both adhere to the surface and maintain appropriate pressure; when opening or closing a cabinet door, it needs to accurately perceive the internal push-pull spring mechanism. Robots need to not only know "where to go" or "where to reach" but also understand "how much force to apply." Before the force-position hybrid control algorithm, these problems had to be solved using force sensors.

With the deep application of the force-position hybrid control algorithm, robots can now perform various operations such as position tracking, force application, force tracking, and compliant interactions without force sensors, significantly improving task success rates. Moreover, it has achieved important breakthroughs in human-robot collaboration safety, laying the foundation for expanding future robot application scenarios.

Relying on this innovative algorithm, the capabilities of robots continue to expand. It can not only accurately perform basic operations such as trajectory execution and force application but also respond promptly to external forces when they are detected. This characteristic also enhances personnel safety during human-robot interactions.

In addition to improving safety, this algorithm significantly enhances the coordination of human-robot collaboration. In scenarios such as jointly moving boxes or pushing carts, the robot can synchronize its movements with humans. It accelerates when humans accelerate and stops quickly when humans stop, adjusting in real time based on the human’s movements. At the same time, the algorithm also exhibits strong generalization capabilities, allowing it to be stably applied even when different types of robots are used.