摘要
The simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) enhances millimeterwave (mmWave) communication by enabling full-space coverage, making it ideal for vehicle-to-everything (V2X) applications. This paper explores a multiuser multiple-input single-output (MU-MISO) mmWave non-orthogonal multiple access (NOMA) downlink vehicular environment aided by STAR-RIS, aiming to maximize the sum rate of the infrastructure-to-vehicle (I2V) links through the design of base station (BS) beamforming, power allocation, and STAR-RIS phase shifts under coupled phase constraints. We propose an unsupervised learning model for STAR-RIS phase design, complemented by analytical approaches for BS beamforming and power allocation. Simulations using the simulation of urban mobility (SUMO) software confirm the superior performance of the proposed scheme in various scenarios while meeting STAR-RIS and NOMA requirements.