Humanoid robots are becoming increasingly capable of walking, handling objects and performing complex movements. But their biggest test may not be what they can do on a demonstration stage. It is whether they can perform useful tasks reliably in real factories.
That is the challenge that Guangdong-based robotics company Topstar is seeking to address with its latest industrial humanoid robot, Star.
Unveiled at the company's TOPS DAY 2026 event in Dongguan, Star is a full-size bipedal robot standing 172 centimeters tall and weighing 72 kilograms. Its two arms have a combined payload capacity of 10 kilograms and are designed for tasks such as material handling and sorting.
Rather than treating Star as a standalone product, Topstar is positioning it as part of a broader push to bring embodied AI into manufacturing. The technology enables robots to perceive their surroundings, make decisions and carry out physical tasks through interaction with the real world.
Topstar has been developing robots in different forms to address a range of industrial needs. Its wheeled humanoid robot Xiaotuo, for example, is being tested in manufacturing applications, including material handling in electronics manufacturing, as well as loading and unloading tasks on production lines. The company has also introduced a quadruped robot, Xingzai, alongside dual-arm systems and intelligent workstations.
But developing robots for different tasks is only part of the challenge. Equally important is enabling them to learn from real-world operations and improve their performance over time.
To support this process, Topstar has developed a robot data collection system that connects several stages, from manual demonstrations and data processing to model training, robot deployment and performance feedback. By collecting information such as images, motion trajectories and gripper actions during real-world operations, the system is designed to support the continuous improvement of robot performance.
The company says its robotics team tested more than 30 scenarios and tasks over the past year, identifying three main areas for further development: sorting small batches of varied products, performing multiple tasks at a single workstation, and intelligent material handling.
These are practical challenges for manufacturers, particularly in production environments where product types and workflows can change frequently. Robots that can adapt to such conditions could help automate tasks that are difficult to handle with conventional fixed automation systems.
The push is taking place against the backdrop of Guangdong's extensive robotics industry. In 2025, the province produced 336,300 industrial robots, accounting for 43.5 percent of China's total and ranking first nationwide for the sixth consecutive year. It also produced 15.18 million service robots, representing 81.7 percent of the national total.
For companies such as Topstar, access to real-world production environments is an important part of developing embodied AI. Its approach brings together industrial scenarios, robots, operational data and AI, with the aim of using experience from factory floors to improve the next generation of machines.
Whether humanoid robots can become a dependable part of everyday manufacturing will depend on more than their ability to move like people. Their real value will be measured by how consistently they can get the job done.
Reporter: Li Muzi
Video: Topstar