Computer Vision aided Beamforming for V2X through Effective Communication

The effectiveness problem proposed by Weaver asks how effectively the received meaning in a communication system affects the goal of the mission. This paper investigates an effectiveness problem that may happen in vehicle-to-everything (V2X) communications. As beamforming is critical for the performance of V2X communication but difficult to implement due to the fast-moving vehicles, computer vision aided beam-forming solves this problem by utilizing computer vision to locate the vehicle for beam pattern selection. With the camera set up at the road-side, the captured image is transmitted to the base station (BS) through a wireless link. The goal of this link between the camera and the base station is not to faithfully transmit of the captured images, but to convey a meaning which allows the base station to find the optimal beamforming. To design this wireless link to solve the effectiveness problem, we leverage the deep learning approach to jointly design the semantic-channel encoder and the effectiveness decoder. Simulations using the ViWi dataset shows the effectiveness of the proposed deep neural network-based codec for the camera-BS link for finding the optimal beam pattern through computer vision.