A Novel Joint AI and Communication Framework with CSI Enhancement

With the development of 6G and Massive Multiple-Input Multiple-Output (Massive MIMO), Artificial Intelligence (AI) and communication has become a significant research focus. This paper designs a novel joint AI and communication framework that integrates dataset generation, deep learning model training/ inference, and communication performance evaluation. The simulation platform has been validated to align with the IMT-2020 calibration results submitted by multiple companies, demonstrating its accuracy in representing real-world scenarios and the credibility of its outcomes. The effectiveness of framework in Channel State Information (CSI) enhancement is assessed through System-Level Simulation (SLS), the results show that while model accuracy decreases with lower compression ratios, the communication performance remains stable within certain limits. This demonstrates the importance of evaluating AI models within the context of communication systems, providing a foundation for future 6G and native AI integration.