Researchers at King’s College London are developing a new wireless networking architecture designed to make future 6G systems more efficient by enabling networks to understand and respond to application goals rather than simply transporting data.

The project, called GoAgentNet, uses a semantic networking approach, meaning the network focuses on the intent behind data rather than treating all traffic equally. The team said this could allow applications and network infrastructure to coordinate in real time, improving performance while reducing bandwidth demands.

Today’s wireless networks typically separate applications from the underlying infrastructure, with networks acting largely as data pipes while applications make decisions independently. This creates inefficiencies as networks struggle to prioritize resources when supporting growing numbers of connected devices and latency-sensitive workloads.

GoAgentNet introduces two new virtual layers, a "knowledge layer" and an "agent layer," between the traditional application and network layers. The knowledge layer gathers task-related information, while the agent layer coordinates AI-driven actions across devices and infrastructure to meet user-defined goals.

In one example, remote robotics systems could exchange only essential goal-related information instead of continuously transmitting large volumes of video and sensor data. The researchers said this could reduce network load while enabling faster coordination between devices.

In the announcement, project lead Yansha Deng said today’s remote robotics systems rely on streaming large volumes of video and sensor data, which limits scalability. Deng said the platform allows users to specify a goal, with AI agents breaking the task into smaller steps so devices exchange only essential goal-oriented information.

The architecture also incorporates edge computing, allowing computational workloads to shift dynamically between devices and centralized infrastructure depending on network conditions. According to the team, this could help maintain service quality even when connectivity is limited.

The work is supported by a $3.5 million Horizon Europe grant and involves a consortium of academic and industry partners, including Nokia Bell Labs and the University of Padova. The research builds on earlier demonstrations of goal-oriented communication shown at Mobile World Congress in 2025, including projects focused on warehouse automation and weather forecasting.

While 6G technology standards remains in early development, the team argues that future networks will need to move beyond traditional packet-based designs to meet expectations around low latency, energy efficiency and large-scale automation. By combining semantic networking with multi-agent AI systems, the KCL team is aiming to create networks that prioritize outcomes rather than raw data throughput.

Deng said the goal is to create a “universal framework that puts the user first,” combining goal-oriented networking with edge computing to support applications ranging from smart cities to industrial automation.

Future work will focus on testing the architecture in smart cities and industrial environments, where real-time coordination between devices and network infrastructure is increasingly critical.