A complete ecosystem of embodied AI in one place
During the World Robot Conference 2026, X Square Robot from Shenzhen presented a complete range of embodied AI technologies, focusing on the integration of basic models, robot hardware, manipulation systems and data infrastructure. The exhibition took place at booth C107 in Beijing from August 19 to 23. The presentation was not limited to just one application - the company showed how its approach can be scaled between industry and home.
At the heart of the exhibition was the WALL-B model, which serves as a foundation for many applications. With self-developing arms with six degrees of freedom and an advanced perception system, this model allows for the implementation of complex tasks in physically changing conditions - from moving packages to arranging flowers. This is not just a technical demonstration, but a sample of the company's strategy: building intelligence that works in the real world, and not just in a controlled laboratory.
Logistics as a test of real-world operation
One of the most concrete examples was a logistics demonstration conducted before WRC 2026. During the livestream, the X Square Robot system processed 1816 packages in one hour, achieving over 98 percent accuracy. This was not just automation - it was about the ability to make decisions in real time. The packages came in different shapes, sizes and arrangements; some were crushed, others had reversed labels.
The system using the WALL-B model analyzed each package individually. The arms could move it, rotate it, unfold it and place it in the right direction on the conveyor belt. When one of the arms noticed that a package was heading to the wrong lane, it intervened automatically - which shows the ability to handle exceptions without the need for full reprogramming.
Dexterous manipulation in a home environment
Enlarged imageClose zoomPrevious imageAs part of WRC 2026, the company showed how embodied AI can work outside the factory. One of the most intriguing examples was a bouquet arrangement demonstration. A robot with two arms received a language instruction: 'Make a bouquet of red roses, add green leaves and place it in a vase'.
The system had to recognize the color, find the right flower, move it, arrange the vase, add other elements - all in changing conditions. When the demonstrator moved the vase or changed the arrangement of the flowers, the robot did not stop, but adjusted the sequence of actions. This shows the difference between sequential programming and true embodied intelligence - the ability to adapt and understand context.
Data collection as a foundation for AI
Underneath all the applications lies the QUANXTA Zero system - a platform for collecting data for embodied AI. The set includes wearable hardware and manual grippers, which allow recording human movements without physical connection to the robot.
The company claims that the system synchronizes data streams with an accuracy of 1 millisecond and can collect visual, tactile and audio data together with millimeter-level positioning precision. Moreover, X Square Robot claims that 1000 samples collected without a robot and 100 from real robots can give a comparable effect to 1000 samples from real robots - which suggests a significant reduction in costs and time in the training process.
The future of the home as an environment for AI
As part of the "X Family Member Program", the company showed a realistic model of a day at home - from breakfast to dinner and leaving the apartment. Instead of presenting individual functions, the demonstration combined various scenarios: the robot helped in preparing food, answered questions via the application, reminded about leaving and controlled household appliances.
This is not just technology - it is a sample of the future, where robots are not isolated, but operate as part of an ecosystem. However, the difficulty lies in the fact that homes are unstable environments: objects change, people move around, and situations are almost always unpredictable. Therefore, the ability to adapt and understand context becomes key - and this is precisely the goal of embodied AI.
