Shanghai hosts the International Embodied Intelligence Skills Competition, with over 800 teams undergoing black-box blind testing across 23 events

No script, no rehearsal possible; over a thousand robots and more than 800 teams from across the country gathered at the Shanghai Zhangjiang Science Hall, focusing on 23 competition events including manufacturing, inspection and analysis, and warehousing and logistics, launching a three-day "black-box blind test" real-world major exam. Yesterday (October 23), the 2026 Global Developer Pioneer Conference and International Embodied Intelligence Skills Competition officially opened. Hundreds of scenario providers brought out various scenarios with 100% realism, expecting robot hardware companies to solve the problems perfectly. Shanghai is precisely the "matchmaker" of this two-way convergence. Embodied intelligence really does work. Shanghai is committed to being "the city that best understands developers" and is also widely recognized in the industry as "the city with the fastest mass production of embodied intelligence"—with Zhangjiang as the core, a robot can be hand-built within 150 kilometers. However, unlike display-and-performance robots that love backflips, Shanghai's embodied intelligence focuses on "labor is most glorious." It is reported that in June this year, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission of the State Council jointly issued a notice launching the "Humanoid Robots and Embodied Intelligence Real-Scenario Training Special Action," aimed at accelerating the construction of a closed loop of "real-scenario training—data accumulation—product iteration—large-scale deployment." This special action directly targets the bottleneck of embodied intelligence having ever-rising financing amounts yet being "unable to enter factories or do work," and also coincides with Shanghai's embodied intelligence belief in "going into factories to work." This competition became the first implementation event nationwide after the special action was released. Zhong Junhao, Secretary-General of the Shanghai Artificial Intelligence Industry Association, introduced that this competition set up 9 major tracks and 23 events. Previously, the competition solicited the "most painful pain points" from more than a thousand scenario providers nationwide, among which "inspection" and "loading/unloading and stacking" became the two most urgent scenarios. To allow robot hardware companies to fully demonstrate their superb skills, the competition successfully invited scenario providers such as Yanfeng Automotive, the Shanghai History Museum, the Shanghai Institute of Quality Inspection and Technical Research, Bailian Group, STO Express, China Life Property & Casualty Insurance, Tianyi Digital Life Technology, and Goertek, who showed great sincerity by bringing the real needs of production lines and the service industry directly into the competition venue. Embodied intelligence enters real scenarios. Embodied intelligence enters real scenarios. Thus, at the competition site there was a 1:1 replica of the State Grid ultra-high-voltage substation, a retail scenario with lighting brightness, shelf heights, and aisle spacing exactly like real supermarket stores, and even expensive and precise quality inspection equipment was brought in for robots to climb in complex, high-pressure, oil-heavy working conditions, arrange goods among cluttered shelves, and precisely grasp among test tubes of different sizes. "Scenario providers have stated that the top three teams in each event can connect directly with the scenario providers, and are expected to quickly enter actual production lines for on-site testing. This means the probability of winning an award at the competition and thus obtaining orders increases significantly, which is extremely attractive to robot hardware companies," said Zhong Junhao. But the competition venue is also cruel. The competition adopts a "black-box blind test" system, meaning that during the competition materials can be randomly changed and fault points randomly set; robots need to read information on site and autonomously adjust their actions to complete tasks such as grasping, carrying, and assembly. In addition, the scenario providers' greatest concerns—operational success rate, efficiency improvement rate, safety and reliability, and economic feasibility—all have on-site measured data available for inspection. Daxiao Robotics has truly entered Shanghai gas stations to work. It can be seen that the competition has become the best testing ground for the gap between "practice" and "use"—does a robot well trained in the laboratory need to learn from scratch when switched to another production line or another set of shelves? Top students are naturally not afraid of testing. The National and Local Joint Humanoid Robot Innovation Center, located in Zhangjiang, brought its Linglong H, Linglong Biped L and other models, and signed up for 4 events in one go: loading/unloading and carrying, precision assembly, firefighting and extinguishing, and supermarket services. Among them, Linglong H, after lightweight deployment, can handle high-difficulty generalized assembly tasks such as battery assembly and USB plugging and unplugging at the competition site. The goal of the National and Local Joint Center is to win more than one championship at the competition. It was also introduced that Shanghai, with its own appeal, gathered more than 1,200 scenario providers and robot hardware companies, and also plans through the competition to collect standardized, reusable "skill packages" for the robotics industry. So-called skill packages mean packaging the various skills of embodied intelligence into modules. For example, grasping objects, visually recognizing parts, tightening screws, and obstacle-avoidance walking are each individual skill packages; when embodied intelligence executes different tasks, it only needs to call them as needed. For example, if a factory wants to do precision assembly, it can call skill packages such as "visual recognition + grasping + screw tightening" for rapid combination. If switched to a warehouse sorting scenario, it only needs to call the skill packages "walking obstacle avoidance + item recognition + carrying and stacking." This free combination avoids developing a complete robot program from scratch for each industry, greatly improving deployment efficiency. This also means that the embodied intelligence body has become an "Apple" smartphone, and various skill packages are like Apps or mini-programs. Different phone users can choose and install Apps or mini-programs as needed to execute various tasks. Competition site.