Large Language Model-Driven Curriculum Design for Mobile Networks

This paper proposes a novel framework that leverages large language models(LLMs) to automate curriculum design, thereby enhancing the application ofreinforcement learning (RL) in mobile networks. As mobile networks evolvetowards the 6G era, managing their increasing complexity and dynamic natureposes significant challenges. Conventional RL approaches often suffer from slowconvergence and poor generalization due to conflicting objectives and the largestate and action spaces associated with mobile networks. To address theseshortcomings, we introduce curriculum learning, a method that systematicallyexposes the RL agent to progressively challenging tasks, improving convergenceand generalization. However, curriculum design typically requires extensivedomain knowledge and manual human effort. Our framework mitigates this byutilizing the generative capabilities of LLMs to automate the curriculum designprocess, significantly reducing human effort while improving the RL agent'sconvergence and performance. We deploy our approach within a simulated mobilenetwork environment and demonstrate improved RL convergence rates,generalization to unseen scenarios, and overall performance enhancements. As acase study, we consider autonomous coordination and user association in mobilenetworks. Our obtained results highlight the potential of combining LLM-basedcurriculum generation with RL for managing next-generation wireless networks,marking a significant step towards fully autonomous network operations.