A Chinese research team on Wednesday launched a new embodied intelligence platform named "Hui Si Kai Wu," creating a universal brain for different types of robots to perform a wide variety of tasks in multiple scenarios.
The platform was developed by the National and Local Co-built Embodied AI Robotics Innovation Center, also known as the Beijing Innovation Center of Humanoid Robotics.
Traditionally, robotic software systems were developed for a specific task and single-use scenario. The new platform breaks from that model by offering a more versatile and generalized solution for different robots and scenarios.
In the demonstration video released by the center, the new embodied intelligence platform can apply to different types of robots, including robotic arms, wheeled robots and humanoid robots. With the help of the platform, the robots can fully understand the task, execute it properly and handle complex missions across multiple scenarios.
"We can provide a complete solution for a specific vertical market, which includes our 'Tien Kung' humanoid robot and our 'Kai Wu' platform. This solution can, to some extent, form a closed loop to complete the tasks that users need. Additionally, our 'Kai Wu' platform can also exist independently. Robotic software developers and systems integrators can use our platform, choose a robot from the market, and then deploy the robot into the real world to perform the corresponding tasks based on our approach," said Tang Jian, CTO of the center.
The "Tien Kung" robot, developed by the center, was unveiled last April as the world's first full-size, fully electric-driven humanoid robot. In November, it was made open-source, with the aim of promoting the secondary development of humanoid robots and accelerating their integration into everyday life.
China launches new embodied intelligence platform to empower robots
Honor's humanoid robot, Lightning, which swept the 2026 Beijing E-Town Humanoid Robot Half-Marathon on Sunday, is a natural extension of years of accumulation in consumer electronics technology, said its developers.
A leading smart device provider in China, Honor independently developed the model, which dominated the podium at the event as it was used by all three teams whose autonomous navigating robots ran the fastest times.
At the Honor factory in Pingshan District in Shenzhen City, south China's Guangdong Province, where robotics engineers developed Lightning. They said the robot's body design incorporates a simulation system that, through artificial intelligence algorithms, can iterate nearly 30,000 design schemes of varying sizes over three months. Complete and mature systems are also in place for battery, communication, and reliability verification.
"We built a simulation lab from scratch. For the robots, we digitize the entire design and put it into a computer. We have our own material library, which can meet the force, thermal, and chemical property demands for each component, under different environments and speeds. We've accumulated about 1000 kinds of materials. For example, if there's a risk with the robot's neck, we just need to change the material code from 001 to 002. Now, through our simulations, we only need one day to perform parallel calculations on 10 different designs, before creating a mold and verifying it in the lab," said Li Zheng, a senior engineer at Honor.
An autonomous robot capable of completing a half-marathon involves a complete industry chain, with core components including high-precision sensors, LiDAR, motors, operating systems, and control algorithms. The development of robotic marathoners have driven an increasing number of component enterprises to get involved.
Manifold, a tech firm established by newly-graduated PhDs, has developed a 3D spatial memory module, which can model an environment in real time and transform it into images that robots can understand. They said several robots running the half-marathon this year adopted their solution.
"Our device can operate within a one-kilometer tunnel with an error margin of only tens of centimeters. For robots, especially in the absence of GPS, this allows them to accurately determine their location. The underlying technology is a multi-sensor fusion technology that we developed in-house," said Qin Youming, CEO and founder of Manifold.
The Beijing Humanoid Robotics Innovation Center set up a training camp for the marathon event. Many university students came a month ahead of the event to develop and debug their technologies and algorithms based on open-source robot bodies, databases, and training platforms.
"These high-quality databases and highly open-source control algorithms are actually very helpful to us. We no longer need to build the house from the ground up, but can skip the most basic part," said Sun Jingyu, a student from Shandong University.
"Through this racing event, I believe we can make our robots more reliable and stable, while also supporting high-dynamic, high-load movements. This is crucial for robots' future application in both industrial, commercial and domestic scenarios," said Guo Yijie, head of the innovative humanoid department and the Marathon project of Beijing Humanoid Robot Innovation Center.
Engineers share development story behind Beijing humanoid half-marathon champion model
