RIS-Assisted Unsupervised Beamforming in Internet of Vehicles

Abstract

Internet of vehicles (IoVs) will require massive high data rate connections with the base station (BS) to provide promising vehicular entertainment services, such as autonomous driving and traffic management. However, vehicles often encounter obstruction from buildings while traveling in urban areas, resulting in blocked direct links between the BS and the vehicle, thereby impacting the channel quality of vehicular links. To enhance the channel capacity of the IoV, reconfigurable intelligent surface (RIS) technology is introduced to assist vehicular networking scenarios in improving the signal propagation environment. Firstly, considering the maximum transmit power budget of the BS and the phase shift constraints of the RIS, we formulate a non-convex optimization problem to maximize the channel capacity of the vehicle-to-infrastructure (V2I) links by jointly designing the active beamforming at the BS and the passive beamforming matrix at the RIS. Then, we propose a RIS-assisted unsupervised beamforming (RAUB) algorithm with a two-phase transformer network architecture to design the active beamforming and the reflection phase shift. By utilizing the global feature extraction ability of the transformer network model in the two-phase network architecture, the learning performance of the network is further improved. Simulation results demonstrate that compared with the traditional block coordinate descent (BCD) algorithm based on alternating iterative optimization, the proposed RAUB algorithm can achieve comparable performance for the sum V2I capacity with lower computational complexity and better convergence performance.