Intelligent Predictive Beamforming for Integrated Sensing and Communication Based Vehicular-to-Infrastructure Systems

Integrated Sensing and Communication (ISAC) has become a promising paradigm for next-generation wireless communications, which are capable of jointly performing sensing and communication operations. In ISAC systems, sensing accuracy and transmission rate are two major metrics to be targeted. In this paper, we propose a predictive beamforming approach based on the multi-dimensional feature extraction network (MDFEN) for vehicle-to-infrastructure (V2I) systems. In particular, in order to achieve high precision and low latency beamforming, the roadside unit (RSU) will perform angle parameter estimation and prediction based on the ISAC signal echoes. Furthermore, our predictive beamforming approach based on the multi-dimensional feature extraction network (MDFEN) is capable of improving the efficient beam alignment by exploiting the joint spatio-temporal characteristics of the received signals at the RSU side. Simulation results demonstrate that the proposed approach achieves a higher accuracy in angle tracking compared to convolutional neural network and long short-term memory models. At the same time, the system is capable of obtaining a higher transmission rate.