Researchers at the Hong Kong University of Science and Technology (HKUST) have proposed a communication concept for future 7G networks, designed to handle data exchange among millions of autonomous AI agents. Under the proposal, the network itself would decide when and to whom to send information, rather than simply relaying all traffic. The study is published in the journal npj Wireless Technology.
Today’s communication networks were built for one job: moving as much data as possible, as fast and reliably as possible. But the rise of smart cities, self-driving cars and digital twins now means constant data exchange not between people, but among millions of autonomous systems at once.
Simply adding more bandwidth won’t fix that, the researchers argue, according to TechXplore. Without new approaches, the communication infrastructure itself could become a bottleneck. That would slow the growth of collective machine intelligence—networks of AI agents that coordinate actions without human input.
To avoid such an overload, the team proposed a different interaction model. Before sending a message, an AI agent would first assess its value and identify the right recipient. It would then need to predict how that information will affect future decisions across the network.
Підписуйтесь на Mediasat у Telegram: тут найцікавіші новини ТБ та телекому
The researchers describe this as a three-stage cognitive cycle. First, the system turns a human goal into a concrete task for the network. Next, an algorithm picks the optimal strategy, balancing bandwidth, latency and signal strength.
In the third stage, the network draws on an internal model of its surroundings to anticipate change ahead of time. The AI needs to track a user’s movements and shifts in the environment, adjusting its behavior before connection quality actually drops.
6G developers are already exploring similar territory, building networks that read the environment around a user rather than just carry a signal. The Hong Kong concept goes further. Under the proposal, self-driving cars would exchange only the data needed to prevent a crash, while factory robots could coordinate repairs before a breakdown even happens. The network stops being a mere data pipe and becomes a platform for collective machine reasoning.
The authors stress that the work isn’t a finished technical standard, but a roadmap for future research. Questions around trust in AI and the stability of these new systems remain unresolved.
