Robotics researchers and industry executives told the 2026 one6G Summit in London that today's wireless protocols cannot support physical AI, arguing that machines operating in the real world — from self-driving cars to factory robots and surgical tools — will need 6G-grade networks and fresh spectrum to function safely.

Lorenzo Masia, executive director of the Munich Institute of Robotics and Machine Intelligence, was blunt about the shortfall. "The protocols we use are not robust enough," he said.

His complaint was not about latency but about scale. "It's more a matter of when you have multiples of these robotic agents and human agents that need to communicate in real time," he explained. "You need something that can handle large numbers. If you're using Bluetooth, forget it. Nothing works. Wi-Fi works a bit better. Another protocol in terms of communication would be much better."

Robots that cannot rely on the network

Masia's speciality is wearable robots: exoskeletons and similar devices designed to improve or enhance human mobility. AI matters here because the machines must react to the wearer's inputs in real time — something static, pre-programmed instruction sets cannot deliver. Exoskeletons could find use in healthcare, helping people with mobility issues. Researchers at the conference were more skeptical about defense applications, which several still described as "science fiction".

Safety requirements dominated the presentation from Francesco Ferro, CEO of Spanish humanoid-robot maker PAL Robotics and president of euRobotics. Public mobile networks, he argued, are too vulnerable to carry the load on their own.

"At the moment, our experience is that, when it comes to reliability, we cannot trust the communication side 100%," Ferro said. "The safety that we trust is something that is onboard."

That distinction — safety logic residing on the device rather than the network — frames a hard requirement for whoever builds the networks that physical AI will run on. Robots may tolerate imperfect connectivity for coordination tasks, but not for the decisions that keep humans out of danger.

Spectrum asks from Huawei's CTO

Wen Tong, CTO of Huawei Wireless, put specific spectrum demands on the table. He claimed the upper 6 GHz band is appropriate for robotics applications, and that the 7-15 GHz band could support sensing of human movements as a way to train machines.

"This is very useful because it's repetitive work," Tong said. "If we need to move 1,000 boxes, all you need is a human to do the example and the rest the robot can do."

The claim carries commercial weight. Upper 6 GHz is already the most contested spectrum real estate in global mobile policy, with mobile operators and Wi-Fi interests fighting over whether to identify it for licensed 5G and 6G use or unlicensed access. A vendor argument that robotics needs the band for licensed cellular adds a new constituency to that debate — though it remains a marketing-adjacent position from a company with a clear stake in spectrum being assigned to cellular. Sensing in the 7-15 GHz range, meanwhile, aligns with the integrated sensing and communication work items already in early 6G standardization discussions.

Cutting the cables in soft robotics

Kaspar Althoefer, director of the Centre for Advanced Robotics at Queen Mary University of London, brought a more tangible problem: cables. His speciality is soft robotics — machines built from compliant materials such as silicone, with designs inspired by animals like octopuses and potential medical uses including surgery.

The device he presented carries tactile sensors distributed across its body, but those sensors currently transmit data through physical cables. That, he said, creates cumbersome engineering constraints.

Next-generation wireless such as 6G could solve the problem. "If we could provide smart sensors with structures that have the capabilities of sensing and telecommunications built in, that would bring this technology forward quite a bit," Althoefer said.

The distances involved are short but meaningful. Integrating sensing and wireless communications would let operators work from 50 to 100 centimetres away from a surgical robot. In industrial and construction settings, going wireless would extend the operating distance to several meters and remove cabling that currently blocks deployment.

"We are struggling with bringing the cables and the fibres through certain structures," Althoefer said. "They're in the way. If one could provide a telecommunication tool that could help us avoid the cabling, that would definitely improve the way we can make soft robotic systems."

A demand signal for 6G planners

For spectrum regulators and 6G standardizers, the summit's message is a demand signal rather than a finished requirement. Robotics researchers want networks that handle dense populations of communicating agents, safety-critical reliability beyond what public networks deliver, and integrated sensing in mid-band spectrum. With 6G standardization work progressing toward a first release expected around the end of the decade, vendors and operators pushing integrated sensing and upper-mid-band coverage now have a new use case — physical AI — to anchor their case.