Customized User Plane Processing via Code Generating AI Agents for Next Generation Mobile Networks

A team of researchers has published a paper on arXiv proposing a novel use of generative AI for next-generation mobile networks. The study, titled 'Customized User Plane Processing via Code Generating AI Agents for Next Generation Mobile Networks,' investigates how AI agents with natural language processing and code generation capabilities can make 6G systems radically more flexible. The core idea is that vertical applications (like IoT or industrial automation) could submit text-based service requests, and an AI agent would generate the code for a new, customized function block to handle user plane traffic. This block would be capable of inspecting and decoding protocol data units (PDUs) and performing specific actions as dictated by the application's needs.

The research specifically examines the code generation problem for creating these on-demand processing blocks. The authors evaluated various factors critical to the accuracy of the generated code, including the selection of the underlying AI model (like GPT-4 or Claude), the design of the prompts given to the agent, and whether providing a code template improves results. Their findings indicate that under suitable conditions, AI agents are indeed capable of generating functional blocks with the desired behavior. This approach moves beyond static, pre-programmed network functions toward a future where networks can generate new capabilities in real-time based on user demand, a key step toward the autonomous and adaptive vision for 6G and beyond.