Abstract
The growing role that artificial intelligence and specifically machinelearning is playing in shaping the future of wireless communications has openedup many new and intriguing research directions. This paper motivates theresearch in the novel direction of \textit{vision-aided wirelesscommunications}, which aims at leveraging visual sensory information intackling wireless communication problems. Like any new research directiondriven by machine learning, obtaining a development dataset poses the first andmost important challenge to vision-aided wireless communications. This paperaddresses this issue by introducing the Vision-Wireless (ViWi) datasetframework. It is developed to be a parametric, systematic, and scalable datageneration framework. It utilizes advanced 3D-modeling and ray-tracingsoftwares to generate high-fidelity synthetic wireless and vision data samplesfor the same scenes. The result is a framework that does not only offer a wayto generate training and testing datasets but helps provide a common ground onwhich the quality of different machine learning-powered solutions could beassessed.