Autoencoders for Compressed Transmission of Vehicular Data

Abstract

V2V(vehicle-to-vehicle) communication has many applications with regard to intelligent transportation systems(ITS), accident prevention, traffic detection, etc. One of the major challenges in V2V communication is the data transmission in high mobility and bandwidth bottlenecks. CAM (cooperative awareness messages), messages are transmitted at a very high rate to update LDM (local dynamic map) to get better traffic awareness. CAM uses broadcast messaging. This could overload the communication network in a dense traffic scenario. In this paper, a data compression framework using the transmission of autoencoder(AE) models in vehicular sensor networks is proposed, where the comparable AE variants are analyzed. The accelerometer and gyrometer data have been used for analysis and the evaluation of the AE variants is performed. The simulation results show the superior performance of the denoising AE. Security is also a concern in transmitting sensor data to other vehicles. The proposed method adds an extra layer of security due to the transmission of AE models in the initialization phase.