Unfazed even with 2,000 robots and 6,000 people connected at once — here's the secret
Huawei and Beijing Unicom presented this case of network support for the second World Humanoid Robot Games at the National Speed Skating Oval in Beijing on Aug. 23.
The first challenge Beijing Unicom and Huawei faced at the event was explosive traffic. At the opening ceremony, about 6,000 people simultaneously connected to the network in a 12,000-seat arena. Because robot competitions are an unfamiliar spectacle, spectators repeatedly shared videos they filmed on-site, pushing per-person data usage 1.5 times higher than usual. Beijing Unicom explained this was "equivalent to about 15,000 people connecting at the same time."
Unlike conventional sporting events, not only spectators but also robots had to use the network simultaneously. A public network for general users and a dedicated network for robots had to be operated stably together. The key performance metrics also differed. Whereas download speed took priority in existing competitions, latency mattered far more than speed for robot control.
In response, the two companies built an indoor network with 300 MHz of bandwidth and applied dedicated robot carriers, network slicing and AI-based operations technology. Robot control latency was reduced to about 18 milliseconds, while public network speed exceeded 160 Mbps (bits per second). Basic service performance improved by more than 30%.
Network's presence grows in just a year… "core infrastructure for the AI era"
The network's standing in the robot industry has changed dramatically in just a year. Yang Lifan, deputy general manager of Beijing Unicom, said, "Just a year ago, the robot industry's demand for networks wasn't that great." At the time, many robots were operated by people controlling them remotely from nearby.
The turning point was the humanoid robot half marathon held in Beijing's Yizhuang in April this year. As robots had to move autonomously on wide roads, the need for a wide-area mobile network grew. All of the top six robots at the time received network support from Beijing Unicom. While some functions can currently be handled by Wi-Fi, the explanation is that it has serious limitations in large spaces such as a marathon.
Battery issues are also increasing dependence on networks. Robots already use a lot of power to move their motors and joints, and adding AI chips to the body increases consumption further. Li Jie, president of Huawei's Wireless TDD Product Line, said, "Many leading robot companies are already moving AI computation to the cloud." Under this approach, the robot body concentrates power on movement, while complex inference is processed in the cloud.
In this case, the network acts as a "nervous system," sending data collected by cameras, lidar and other sensors to the cloud and then delivering control commands back. Li said, "The level of AI intelligence will largely depend on network stability," naming the network as core infrastructure of the AI era.
Future network focus shifts from speed to stability… robots in healthcare by 2030
The core of future network innovation is expected to shift from competing over top speeds to stability. For robots and AI terminals to send large-scale training data to the cloud, uplink guarantee standards must be raised, and to save battery power, a collaborative framework is needed so that networks and terminals operate efficiently only when necessary.
The problem is cost. As with these games, stable latency of about 20 to 30 milliseconds can be guaranteed in a limited-range arena, but applying the same approach across all of Beijing would sharply raise costs. Huawei proposed intelligent operations as the solution — using AI to identify and predict problems in advance and deploying resources only where needed.
As networks advance, the fields where robots are applied are expected to widen. Huawei cited quality inspection, logistics sorting and healthcare as promising areas around 2030. Li said, "By combining physical AI, large models and highly reliable, low-latency networks, we will be able to solve far more complex problems than we can now."
