In 2025 WAIC, Galaxy general-purpose robots showcase their multi-scenario skills collectively.

As one of the most influential events in the global artificial intelligence field, the 2025 World Artificial Intelligence Conference (WAIC) grandly opened in Shanghai on July 26.

After the opening ceremony of the conference, Li Qiang, the Premier of the State Council, toured the exhibition halls together with foreign guests and representatives of international organizations in attendance. In front of the booth of Galaxy Universal, Galbot fetched and delivered products from the shelves to Premier Li Qiang, completing a real - world demonstration of autonomous reasoning and zero - remote - control operation.

Behind this outstanding performance is the end - to - end embodied large model independently developed by Galaxy Universal. With its leading technological strength, Galaxy Universal became a shining star in the technology field at the conference. As an industry leader, it demonstrated the vigorous development and wide influence of China's embodied intelligence industry.

The VLA large model drives robot productivity, and Galaxy Universal's Galbot is rated as the "treasure of the exhibition" at the conference

At this year's WAIC, Galaxy Universal was invited to participate in the highly anticipated "2025 Skills Stage". Inside the exhibition area, Galbot carried out all - day supermarket operation work based on the team's self - developed end - to - end embodied large model, GroceryVLA.

During the leaders' tour of the exhibition halls, Galbot accurately fetched and delivered products for Premier Li Qiang, foreign guests, and representatives of international organizations. In front of the booth, Galbot worked stably for a long time without interruption, continuously demonstrating its excellent performance and reliability. It vividly illustrated how embodied intelligence technology can help the retail industry move towards intelligence, winning high praise from the leaders and guests on - site.

Galbot's continuous and stable performance at the Skills Stage fully reflects Galaxy Universal's solid steps in transforming cutting - edge technologies into new quality productivity. Based on this innovative practice, this exhibition area was rated as the "treasure of the exhibition" at the 2025 WAIC by the conference organizing committee and became a popular landmark for professional visitors to check in.

Multiple capabilities are showcased for the first time, constructing a closed - loop for embodied intelligence scenario applications

At its self - organized booth at this year's WAIC, Galaxy Universal comprehensively demonstrated the core advantages and diverse application scenarios of the robot industry, covering multiple fields such as retail, industry, logistics, and urban services.

Commercial retail: Supermarket - guarding robots driven by GroceryVLA

The "Skills Stage" exhibition area was very popular. Galaxy Universal also built a 1:1 restored real supermarket scenario in its self - organized exhibition area, where Galbot guarded around the clock for the demonstration to the guests and visitors at the conference.

Different from traditional display racks, the shelves brought by Galaxy Universal on - site had densely arranged products with a wide variety of SKUs, and the products were arranged in various forms, including bagged snacks hanging on hooks and tightly arranged soft - packed drinks in complex layouts.

The key reason why Galbot can operate stably in such a complex environment lies in the end - to - end embodied large model behind it, GroceryVLA. This model breaks through the traditional separate design of "vision + trajectory planning". It can autonomously identify products and stably grasp them in real - world scenarios with almost no gaps, tightly arranged shelves, and a large number of SKUs. It can operate flexibly and efficiently without path planning, indicating that the closed - loop control driven by the end - to - end large model has commercial viability.

This model does not require separate parameter adjustment for different product packages. It supports the precise grasping of various forms of products, such as soft - packed (e.g., bagged bread, marinated eggs), hard boxes, plastic bottles, and transparent jelly cups, achieving a unified grasping strategy across categories. It can accurately pick and place both rigid and flexible objects, meeting the complex needs of full - category retail.

Dr. Wang He gave a technical explanation at the booth.

In the multi - scenario deployment at this year's conference, in the face of challenges such as complex lighting changes and random product placement, the model still maintained stable and efficient performance. Throughout the deployment and formal operation period, Galbot achieved full - process operation without remote control and without pre - collection of scenario data, fully demonstrating the model's strong generalization ability.

Industrial manufacturing: Robots complete SPS sorting work

In the SPS sorting work of automobile parts in the industrial field, Galbot easily overcomes the complex challenges that traditional robots struggle with. Multi - layer misaligned stacked parts, workpieces with similar appearances, and various environmental interferences greatly test the robot's recognition and grasping abilities. Galbot can not only accurately identify and grasp the target workpieces but also distinguish similar parts to avoid mis - grasping. It can autonomously avoid obstacles during deep - basket operations and complete intelligent operations.

What's more remarkable is that in the face of human interference, such as swapping the positions of workpiece baskets or deliberately messing up the workpieces, Galbot can still quickly locate the correct workpieces, showing high - level autonomous decision - making and environmental adaptation abilities. With its excellent generalization ability, Galbot has reached a cooperation agreement with an internationally renowned automobile enterprise and conducted application - landing tests, helping automobile manufacturing factories achieve efficient and stable automated sorting.

Industrial materials: Robots complete material box handling work

In the transfer operation of industrial and logistics material boxes, Galbot demonstrates intelligent working abilities. Facing multi - layer stacked material boxes without QR codes or labels for auxiliary positioning, the robot completely relies on visual perception to autonomously identify the position and shape of the boxes and intelligently plan the handling actions. It can flexibly use one hand to open the stacking gap first and then use both hands to cooperate in handling, showing a high degree of autonomous action planning and execution ability.

On - site at the booth, Galbot's actions were smooth and natural, efficient and fluent, winning exclamations from professional visitors. In cooperation with the Geely ZEEKR automobile factory, Galbot's box - handling efficiency has reached a level comparable to that of human workers, fully reflecting its practical application value in complex industrial and logistics scenarios.

Urban services: Intelligent robotic dogs for autonomous garbage collection

In the field of urban services, Galaxy Universal's self - developed model is deployed across different robot bodies, enabling robotic dogs to collect garbage and conduct inspections. Visitors can randomly throw items such as water bottles in the exhibition area, and the position and posture of the items are completely random. The robotic dog can autonomously and accurately identify the target objects and quickly pick them up, fully demonstrating the system's high adaptability to complex dynamic environments and precise operation ability.

Based on the breakthrough results in the "motion - operation integration" field, the team developed the Generalized Oriented Reachability Mapping (GORM) technology, which greatly improves the whole - body operation generalization ability of robots under a six - degree - of - freedom base. Combined with the reinforcement learning strategy, the robot achieved a grasping success rate of up to 89% in dynamic and complex environments. The relevant research results have been accepted by the top international robotics conference, ICRA 2025.

In addition, Galaxy Universal integrated the grasping posture detection and motion planning modules. Through modular design, it efficiently connects motion, perception, operation, and planning, enabling seamless cooperation of robots in multiple scenarios. Based on this architecture, robots can not only handle diverse tasks such as household cleaning and disaster rescue but also flexibly cope with various challenges in urban maintenance.

The rich exhibition of embodied intelligence scenario applications in the exhibition area fully reflects Galaxy Universal's technological leadership and industrialization implementation strength in the field of embodied intelligence, as well as the company's firm determination to build an open and win - win intelligent ecosystem.

At this year's conference, Galaxy Universal won the 2025 WAIC SAIL Award for the outstanding innovation ability of its project "Embodied Base Large Model for Cross - industry Applications Driven by Synthetic Data". This award is known as the "Oscar of artificial intelligence" and is a high - level recognition from the industry for technological innovation and business influence.

This project covers multiple real - world application scenarios such as commerce, industry, and services, demonstrating Galaxy Universal's important breakthroughs in synthetic data training and commercial application of embodied large models, and setting a new benchmark for promoting the cross - industry application of embodied intelligence technology.

Taking the application of embodied intelligence technology in the commercial retail field as an example, Galaxy Universal's world - first intelligent retail solution for humanoid robots has been officially put into application. Based on the self - developed embodied end - to - end large - model technology, Galbot completes the full - process operations such as inventory checking, replenishment, product fetching and delivery, and packaging. It supports 24 - hour uninterrupted product delivery and inventory management, realizing efficient and intelligent operation with robot guarding.

Relying on this solution, new physical stores can be deployed in just two days, showing strong replication ability and promotion potential. The solution is compatible with various complex shelf structures such as open shelves, precision extraction shelves, and drawers. The robot can autonomously and efficiently complete the guarding work, and couriers only need to enter the password to pick up the goods.

Currently, Galaxy Universal has received orders from 100 stores and plans to put them into use nationwide by the end of the year. Nearly ten stores in Beijing have been deployed and are in normal operation. Each store of about 40 square meters is equipped with more than 5,000 kinds of medicines and 6,000 cargo lanes. Tens of thousands of boxes of medicines are managed by Galbot alone, which guards around the clock and operates efficiently and stably.

Sharing insights at authoritative forums, leading the frontier of embodied intelligence technology

In the process of promoting the commercial implementation of embodied intelligence technology, Galaxy Universal not only achieves technological breakthroughs but also actively participates in high - level academic and industrial forums in the industry, demonstrating the company's leadership in the field of cutting - edge technologies.

Dr. Wang He, the founder of Galaxy Universal, was invited to participate in multiple important forums during the 2025 World Artificial Intelligence Conference and shared a number of cutting - edge technological achievements as an expert. On July 26, as an elected member of the MIT TR35 young scientists, Dr. Wang He was invited to attend the "AI Future Development Forum" and had in - depth exchanges on the theme of "From scientific intelligence to general artificial intelligence".

During the conference, Dr. Wang He was also invited to attend the "2025 Large Model Forum", the "2025 WAIC 'Artificial Intelligence Interdisciplinary Science Forum'", and the "Light Source Artificial Intelligence Industry Integration Forum", sharing his valuable insights on the development of the embodied intelligence field with the guests at the conference.

From scientific research breakthroughs to product implementation, Galaxy Universal always adheres to the development concept of "technology serving real - world scenarios". The multiple robot capabilities showcased at this year's WAIC are not only the display of single - point technological achievements but also a true reflection of its industrialization strategy. By connecting the entire closed - loop chain from model training, system integration to real - world deployment, Galaxy Universal is vigorously promoting the embodied intelligence technology from "show - able" to "truly usable" with strong execution ability.

Facing the future, Galaxy Universal will continue to deepen cooperation with multiple parties such as the government, enterprises, and scientific research institutions, build an open and win - win embodied intelligence ecosystem, further unleash the potential of embodied large models in more industry scenarios, empower all industries, promote the large - scale implementation and industrial application of general robots, and inject strong impetus into China's intelligent manufacturing.