Galbot unveils sporty new humanoid model

Earlier this year, a humanoid robot stole the spotlight at China's Spring Festival Gala, folding clothes alongside actor Shen Teng. Now, the robot has learned a new challenge: playing tennis.

Developed by Galbot, a humanoid robotics company known for its wheeled robots, the machine made a return appearance at the 2026 World Robot Conference, where it was joined by a new sibling, the Galbot ET1. The ET1 marks the company's first bipedal humanoid robot and is being trained to play tennis.

The ET1 is scheduled to take part in the opening performance of the World Humanoid Robot Games on Aug 22, where it will interact with tennis legends and high-profile entertainers.

For Galbot, however, tennis is about far more than putting a robot on the court.

"Tennis is an extremely complex activity," said Wang He, the company's chief technology officer. "It requires precise control of the robot's entire body to maintain balance, while its wrist movements must be highly precise, so the racket makes contact with the ball."

The physical challenges are only part of the problem. A robot must also decide what to do: where to hit the ball, how to get it over the net, and how to score.

In Wang's view, tennis brings together the capabilities of both the cerebrum and the cerebellum, combining high-level decision-making with precise motor control. That makes it a particularly demanding test for humanoid robots and potentially a milestone for the technology.

In 2016, Google's AlphaGo defeated one of the world's top Go players, demonstrating that an artificial intelligence system could master a complex task once thought to require uniquely human intuition. Galbot believes tennis could provide a similar "AlphaGo moment" for humanoid robots.

Behind the ET1 is Galbot's self-developed AstraBrain-Agent, an agent architecture designed for real-time multimodal interaction and motion control in humanoid robots.

The system allows a robot to observe its surroundings continuously, understand spoken instructions, and dynamically plan its actions. Rather than relying solely on prerecorded movements or motion-capture data, the robot can learn new skills through interactions with people and its physical environment.

That approach points to a broader goal for humanoid robotics: moving from robots that repeat predefined actions to robots that can comprehend unfamiliar situations, make decisions, and acquire skills in the real world.

In the conference's exhibition area, the robot also demonstrated a range of household tasks, including putting objects away, washing and hanging clothes, and folding clothes on a bed.

The demonstrations exhibit how the company is seeking to bring humanoid robots closer to everyday life.