The highly anticipated World Robot Conference 2025 recently opened in Beijing. Themed “Making Robots Smarter and Embodied Intelligence More Intelligent,” the conference is one of the world’s most influential events in the robotics industry.
Yongkun Wang, Founder and Chairman of Standard Robots, was invited to speak at a concurrent event, where he delivered a keynote titled “Evolving Through Iteration, Innovating Through Real-World Applications: Standard Robots’ Industrial Foundations and Technology Journey.” Below are some of the key insights from his presentation.
China Will Lead the Development of Embodied AI and General-Purpose Robots for Factory Applications
Since its founding, Standard Robots has remained deeply rooted in the industrial sector. Starting with autonomous mobile robots and building on their ability to understand their surroundings, we have continuously expanded their capabilities by integrating different modules—from early lifting, conveyor, and towing modules to today’s single-arm and dual-arm systems. As a result, our robots can perform an increasingly diverse range of tasks.
I believe China will be the first country to develop embodied AI and general-purpose robots for factory applications. With the world’s most comprehensive industrial supply chains and an exceptionally wide range of real-world use cases, China provides the ideal testing ground for robotics. Real-world applications generate data, and data enables intelligence. Only by combining data-driven insights with a deep understanding of industrial processes can robots evolve into truly intelligent, general-purpose systems.
Focusing on Frontline Manufacturing Applications
We focus on real-world manufacturing environments, creating measurable value for customers by automating manual work instead of pursuing impractical, overly generalized capabilities. Genuine challenges must be identified on the factory floor, not imagined in a laboratory. The key barriers lie in gaining access to industrial environments and developing deep industry Know-How.
Many robotic products are designed around assumptions, only to encounter numerous long-tail challenges when deployed in actual manufacturing operations—issues that can only be uncovered through hands-on experience on the front line.
Our Goal Is to Create “Super Workers” in Factories Through Embodied AI
As one of the early pioneers in the industrial AMR sector, Standard Robots has achieved industry-leading capabilities across many core technologies. However, customers care less about technological leadership itself than about four essential qualities: safety, reliability, intelligence, and efficiency—the foundations of technology that delivers real value.
We ensure safety and reliability through proprietary foundational technologies and the continuous development of a unified core technology platform. Intelligence and efficiency are achieved by accumulating extensive real-world application experience and intelligently coordinating tasks across multiple robots.
The Key Bottleneck in Embodied AI
The primary bottleneck in embodied robotics today lies not in hardware, but in the “intelligent software” that enables robots to make autonomous decisions.
Robotic manipulation fundamentally involves physical interaction with objects, and every physical interaction increases the complexity of the problem exponentially—making it fundamentally different from autonomous driving.
Robots may perform more effectively when they can reason through and execute tasks according to a complete sequence of steps. This is precisely why we introduced a hierarchical architecture.
A More Practical VLA Approach for Industrial Applications
From a technical perspective, we believe end-to-end Vision-Language-Action (VLA) models hold tremendous long-term potential, but they are not currently the only way to address challenges in industrial environments.
Our approach combines a hierarchical “brain and cerebellum” architecture with both individual and swarm intelligence. It provides a more practical, secure, and deployable path that can create value for customers more quickly.
At the individual intelligence level, our goal is to continuously advance each robot from autonomous navigation to mobile manipulation and ultimately embodied intelligence. Using a fast-and-slow system supported by this hierarchical architecture, we combine general-purpose reasoning models with deterministic task execution to ensure reliable and efficient robot performance.
In terms of swarm intelligence, real industrial environments never rely on just one robot. Even a fully developed “super robot” operating in a factory must receive instructions from information systems rather than individual operators. It must also interact with other robots and equipment to complete tasks.
Through RoboVerse, we enable cross-platform fleet coordination and integration with customer systems instead of focusing solely on the general-purpose capabilities of individual robots. This allows us to establish a complete commercial closed loop—from real-world deployment and the data flywheel to intelligence improvement and accelerated mass production—more quickly, enabling robotic intelligence to evolve through large-scale commercial deployment. While the industry currently places too much emphasis on the general capabilities of individual robots, swarm intelligence is becoming increasingly important.
Therefore, I believe the VLA approach best suited to industrial environments is a hierarchical “brain and cerebellum” architecture that combines VLM-powered reasoning with deterministic execution.
The Vision and Purpose Behind RoboVerse
Humanoid robots are an important addition to the robotics landscape, but they are not the only solution. Manufacturing requires robots in a wide range of form factors, along with specialized equipment, to handle different environments and processes.
We believe specialized and general-purpose robots will continue to coexist in industrial settings, with systems from different manufacturers and brands operating side by side. This makes unified protocols and a shared software environment essential for coordinating and orchestrating diverse robots, enabling them to collaborate and complete tasks within the same facility. We want every robot to speak the same language—and that is the vision and purpose behind RoboVerse.
The Four Stages of Intelligent Manufacturing
China’s manufacturing industry must progress through four stages to achieve true intelligence.
1. Workforce Mechanization. A large number of robotics companies—whether specializing in industrial robots or other robotic systems—operate at the foundation of this transformation. Mechanizing the workforce will therefore require the combined efforts of hundreds or even thousands of companies.
2. Software-Driven Management. Every industrial environment requires a unified software platform to coordinate and manage different types of robots, equipment, and information.
3. Factory Digitalization. Software creates structured data; structured data enables continuous data accumulation; and accumulated data ultimately provides the foundation for data-driven intelligence.
4. Intelligent Manufacturing. This final stage can only be achieved after the first three foundations are firmly established.
These four stages form a pyramid, with each layer building on the one below it. Many companies attempt to work from the top down, moving from intelligence toward workforce mechanization. I believe the transformation must proceed from the bottom up. Without solid foundations established at each stage, intelligent manufacturing will remain little more than a castle in the air.
Standard Robots aims to remain focused on the industrial sector and, through four consecutive five-year development plans over the next two decades, become a true leader in China’s intelligent manufacturing—starting with foundational robotics technologies and building a complete technology stack from the ground up. Our vision is to take Standard Robots from China to the world—from China’s Standard to the world’s STANDARD.
