HoST makes every humanoid robot stand up

In some situations, humanoid robots can quickly lose their footing and fall. A new framework is designed to help them get up again quickly.

listen Print view
Man hits a robot to prevent it from getting up.

The Unitree G1 humanoid robot is prevented from standing up by bumps. The robot manages to stand up anyway.

(Image: Tao Huang (Screenshot))

2 min. read

A research team from China and Hong Kong has been thinking about how humanoid robots can stand up again as quickly as possible after falling down. The result is a system called Humanoid Standing-up Control (HoST). The system is based on artificial intelligence and enables humanoid robots to stand up from all possible positions on many surfaces.

In principle, the HoST framework can be used on any bipedal robot, the researchers write in their study "Learning Humanoid Standing-up Control across Diverse Postures", which has been published as a preprint on Arxiv. The researchers trained the HoST system using the "Isaac Gym" simulator (now Isaac Lab) from Nvidia. This is an open-source physics simulation for reinforcement learning (Reinforcement Learning – RL).

Empfohlener redaktioneller Inhalt

Mit Ihrer Zustimmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.

Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.

The scientists used the simulator to teach a two-legged robot to stand up from different positions and on different surfaces by trial and error. For example, they tried to get the robot to stand up from lying on its stomach, on its back, leaning against a wall and lying on a chair. They also had the robot do this on different surfaces such as gravel stones, grass, wood and foam. Slopes were also taken into account. Actions that led to success were rewarded, while others were penalized.

The scientists transferred the HoST system trained in this way to a real humanoid "Unitree G1" robot. The robot was able to stand up again independently from almost all conceivable positions. Even disturbances during standing up, such as knocks, could not prevent it from finding an upright position again.

Videos by heise

The researchers see the HoST system as a robust approach to how humanoid robots can get back on their feet after a fall – and as quickly as possible. This can also help bipedal robots to better navigate unpredictable environments and get back up after falls.

(olb)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.