How can a robot respond within 1 second when an elderly person falls? This Shanghai store uses robot fighting to find a solution
The robot fighting club created by Shenzhen Quanmingxing Robot Technology uses fighting as an entry point to train robots' strength and rapid response capabilities, with the goal of enabling robots to recognize and respond to scenarios such as elderly falls in less than 1 second. Its first Shanghai store, located in Baoshan Smart Bay, is expected to open as soon as November and is poised to become the country's first robot fighting club, offering research and science education during the day and selling spectator tickets and related consumption at night. The club is positioned as a third-party platform not limited to a single brand; manufacturers such as Zhongqing, Zhiyuan, and Unitree can all bring robots to fight and conduct technical validation.
Body · AI translation
Robot fighting Shu Shu photo On the afternoon of October 23, on the fourth floor of the West Bund International Convention and Exhibition Center in the "Future Creation Hall," the "Boxing Star Club" booth participating in this West Bund International Technology Consumer Carnival (WTCC) was staging a robot sparring match, and the robot that won a round made a flashy street-dance hand gesture to "show off." Although the above postures were all controlled by human engineers, the increasingly agile robots were still among the most eye-catching exhibits on site. The reporter observed on site for two consecutive days that the "AI + consumption" picture presented by this WTCC is not a single product display, but a complete chain from technology demonstration to real delivery. Faced with the topic repeatedly mentioned at various AI exhibitions in recent years—"when can robots be used at home"—this year's WTCC showcased an industrial prospect worth looking forward to: from robot fighting to embodied companion robots, to AI office agents and real-time interactive digital humans, artificial intelligence is accelerating from technology demonstration to real delivery. Consumption scenarios are accelerating into reality At the WTCC exhibition site, robots were the focus that drew visitors to stop. Li Yazhou, general manager and founder of Shenzhen Boxing Star Robot Technology Co., Ltd., introduced that the "Boxing Star Robot Fighting Club" built by the company uses robot fighting as an entry point and is exploring building a "mass-market entrance" for the robot industry, one reason being that the public's understanding of the robot industry still has an information gap. "After communicating with many robot host manufacturers, we found that even in large companies with thousands of employees, not everyone has come into contact with robots or operated one," Li Yazhou said. For the public, more than 90% of consumers have not yet come into contact with a real robot. Based on this judgment, the Boxing Star Club chose offline physical contact to eliminate the gap between on-screen information and real experience. Speaking of the reason for founding the robot fighting club, Li Yazhou compared this competitive model to the relationship between cars and F1, and motorcycles and MotoGP, with the goal of building a robot culture industry IP and creating a third-party platform not limited to a single robot brand. "Zhongqing, Zhiyuan, Unitree—all robot manufacturers can bring their own robots to fight on our platform and carry out technical verification." Robot fighting Shu Shu photo Taking elderly-care robots as an example, being able to lift and carry an elderly person requires the robot to have sufficient strength support; when an elderly person suddenly falls or an object is knocked over, the robot needs to complete action recognition and response in less than 1 second, quickly protecting the elderly person rather than remedying afterward. The above capabilities can all be used to train a robot's strength and rapid response ability through fighting competitions, and then be applied to life and commercial scenarios. At present, the first Shanghai store of the Boxing Star Robot Fighting Club, located in Baoshan Smart Bay, is under intense preparation and is expected to open as soon as November, and is expected to become the country's first robot fighting club. Li Yazhou introduced that during the day the club will carry out study tours and science popularization education, and at night it will create a "tech trend spot" providing drinks and light meals, forming a commercial closed loop through selling spectator tickets and supporting consumption. In Li Yazhou's view, the first generation of general-purpose robots that truly enter home scenarios still needs 3 to 5 years of polishing, and the challenges lie not only in technological breakthroughs but also in the lack of real-scenario data and adaptation to physical space. The good news is that the two major industries of robots and home appliances are currently "moving toward each other": home appliance manufacturers have begun to proactively modify products to adapt to robots, for example developing refrigerators with remote automatic door opening; robot manufacturers are focusing on standardized module development, and a major domestic company has already divided household chore scenarios into more than 20 functional modules in the laboratory, training robots one by one to perform operations. Bringing costs down In the home companion robot track, Qingxin Yichuang brought the embodied character robot "Amoo" to the WTCC site, which has achieved deep interaction at the semantic level within half a year of its debut. Users can trigger Amoo's proactive response and emotional soothing behavior through natural language instructions, and autonomous navigation and obstacle avoidance functions have also been added. Supported by the Dino OS foundation model self-developed by Qingxin Yichuang, Amoo can replicate the images and speech tones of different IPs in batches, adapting to scenarios such as early childhood companionship and elderly care. However, even though the application scenarios are already relatively broad, Amoo still faces the common challenge of robots landing on the consumer end. "User research shows that the two challenges consumers are most worried about regarding home robots are safety and trust," introduced Xu Huanqing, CMO of Qingxin Yichuang. Safety refers to whether the robot will fall when moving in the home environment, and whether it can ensure that humans are not injured after falling; trust refers to consideration of user privacy protection. Amoo Shu Shu photo Qingxin Yichuang's approach is to equip Amoo with a plush outer skin covering the whole body and a design without sharp corners, and also adopt power-off protection and fall-and-get-up mechanisms. In terms of data storage, Amoo uses a combined storage technology of local and cloud, interacting with the cloud only when large model computing is involved. Cost is another threshold that robots must cross for commercial use. Xu Huanqing revealed that motors are currently the main source of cost for robots. Taking Amoo as an example, with 25 degrees of freedom, it relies on high-precision motor drive. There are currently two main ways to reduce costs: first, using standardized motors and relying on mass production to lower prices; second, robot companies developing their own motors, which, although requiring large upfront investment, are better suited to their own products later. Qingxin Yichuang adopts a full-stack self-developed technology path including motors. Policy support is also a path to alleviate the cost pressure on robot companies. Xu Huanqing said that the policies enterprises currently look forward to mainly come from three aspects: first, opening up application scenarios, hoping the government will coordinate downstream partners, such as hotels, hospitals, and urban public spaces, to open up the ecosystem; second, providing test site support, hoping to obtain more public or dedicated robot test sites; third, computing power support. As research on the models carried by robots gradually deepens, enterprise computing power costs rise, and it is hoped that the government will continue to implement and optimize policies such as computing power vouchers. More than just home use The landing of AI agents on the consumer end is also worth attention. SenseTime brought multiple exhibits to this WTCC, including the SenseNova large model, the RaccoonWork AI office agent, the Realita real-time video interactive digital human, and explanation robots. Pan Wangqun, senior product operations manager of SenseTime Raccoon, said that what the company most hopes is that visitors coming to a C-end consumer exhibition like WTCC will hand real tasks to the agent, "to see whether 'Raccoon' can really get things done." At the WTCC site, consumers can try using "Raccoon" to organize travel plans, turn scattered materials into clearly structured personal knowledge archives, design processes and material lists for offline events, and even "package" their own experience into a reusable agent. In non-office scenarios, garden drawing reviewers can precipitate years of professional practical experience into an agent to assist humans in checking drawings; history enthusiasts can use agents to organize biographies, historical events, and geographical routes; visually impaired people can also use "Raccoon" to improve information access and the experience of using digital tools. Regarding the current progress of consumer-end AI agents, Pan Wangqun said that AI agents can already handle many tasks with clear boundaries, such as organizing information, analyzing data, planning activities, and calling tools to complete multi-process operations. "But agents have not yet entered the stage of truly 'understanding people.'" She told reporters that before truly understanding human needs, agents still need breakthroughs in at least four aspects—long-term understanding, contextual judgment, proactivity and sense of boundaries, and trustworthy execution. "Evaluating whether AI 'understands users' cannot be reflected only in whether the conversation is natural, but in whether AI can help us complete real tasks in a long-term, stable, and trustworthy manner," Pan Wangqun said. Fan Ling, founder and CEO of Tezign, believes that this year is a watershed, marking that AI agents on the consumer end are advancing from "generating a result" to "completing a task for people." He also pointed out the difference between personal agents and industry-end agents: the former focuses on serving individuals in life scenarios, while the latter focuses on organizational judgment and process building. AI agents also face cost challenges. Fan Ling said that agent development costs can be divided into four parts, namely model invocation, context construction, engineering and operation, and the most easily overlooked "judgment capability." "Not all tasks are worth using the largest model; choosing the right model is the way to effectively reduce the cost of using agents."