Large Beamforming Models for Robust Wireless Transmission in Generalized Scenes of RIS-aided Intelligent IoV Network

In recent years, concomitant with the rapid advancements in large models (LMs) technologies within the field of artificial intelligence (AI), the domain of intelligent Internet of Vehicles (IoV) network has also experienced swift progress. Those potential applications such as autonomous driving, intelligent driver assistance systems, personalized driver experience, and traffic flow and urban mobility management are dependent on the excellent IoV wireless communications system. However, there still exist many fundamental challenges when involving LMs into IoV network. How to provide effective coverage for intervening massive heterogeneous terminals in complex dynamic wireless network environments is a challenging problem. Besides, the generalization and robustness of beamforming within various scenarios are the fundamental for IoV applications mentioned above. To address the aforementioned issues, we in this paper formulate and integrate LMs to achieve excellent hybrid beamforming within IoV wireless communications system which is aided by reconfigurable intelligent surfaces (RISs) and base stations (BSs) to provide additional channels and coverage gains. Benefiting from RISs, wireless channels become complex and reconstructable, thus providing the possibility of creating LMs based hybrid beamforming. The proposed LMs are formulated by utilizing deep reinforcement learning (DRL) to effectively interact with environments so that robust beamforming can be maintained within RIS aided IoV network. Finally, we exploit the excellent performance of LMs enabled hybrid beamforming and investigate the impact of critical learning parameters. Comprehensive experimental results have consistently confirmed that the proposed LMs enabled hybrid beamforming for RIS aided IoV outperforms traditional schemes in the respect of system throughput and network coverage under dynamic wireless environments