A deep learning approach for predicting and optimizing the V2X network parameters for sustainable smart transportation systems

Abstract The increasing prevalence of automated vehicles necessitates robust vehicle-to-everything (V2X) communication systems. However, challenges such as latency, reliability, transmission rates, data collisions, and security vulnerabilities require thorough investigation. To address these issues, this study proposes a deep learning-based hybrid model integrating long short-term memory (LSTM) neural networks and generative adversarial networks (GANs) to predict performance metrics in V2X communication networks. The GANs augment the sample vehicular dataset and facilitate proactive vehicular network management. The proposed approach comprises three key phases: data augmentation, feature extraction, and predictive analysis. To optimize the neural parameters, the whale optimization algorithm (WOA) is utilized and benchmarked against other optimization algorithms, including beetle swarm optimization, artificial bee colony optimization, and pelican optimization. The parameters in the developed hybrid GAN-LSTM model and autoencoder were tuned using the WO algorithm to minimize key performance metrics such as the mean absolute error, root mean square error, bit error rate, and mean square error. Results indicate that the predictive model enhances the sustainability and efficiency of transportation systems by accurately forecasting critical metrics, including latency, throughput, and reliability, within V2X networks.