Vehicular Environment Identification Based on Channel State Information and Deep Learning

This paper presents a novel vehicular environment identification approach based on deep learning. It consists of exploiting the vehicular wireless channel characteristics in the form of channel state information (CSI) in the receiver side of a connected vehicle in order to identify the environment type, where the vehicle is driving along, without the need of implementing specific sensors such as cameras and radar. We consider the environment identification as a classification problem, and we propose a new convolutional neural network (CNN) architecture to deal with it. The estimated CSIs are used as the input feature to train the model. To perform the identification process the model is targeted to be implemented in an autonomous vehicle, connected to a vehicular network (VN). The proposed model is extensively evaluated and compared to related approaches and also to state-of-the-art classification architectures. The experiments showed that our proposed model yields in favorable performance compared to all considered methods