A Novel Optimized AI Based Model for Traffic Prediction in VANET

Abstract

Research into a novel type of wireless network, the vehicle ad hoc network (VANET), has gained an increasing research interest in recent years. By exchanging data about things like vehicle velocity, location, and direction, VANET facilitates interaction between moving vehicles and fixed infrastructure. It's best to take an alternate route, if possible, when there's a high probability that a large number of cars will use the same one to get from point A to point B. If vehicles could precisely foresee traffic congestion, they could potentially evade it. Hence, suggested using DRL in VANET to improve traffic congestion forecasting. In addition, this deep reinforcement learning model examines and compares various neural networks, such as the CNN, MLP, and LSTM, to determine which is best with the results. The proposed strategy is evaluated based on the ATTD and the AWTD of vehicles, which serve as metrics to assess its effectiveness. Through simulation analysis, it is evident that the suggested approach enhances the average travel time delay and average waiting time delay across multiple iterations, considering the environmental factors as inputs. The simulation results demonstrate that compared to existing algorithms used for network traffic prediction, the deep reinforcement learning model provides superior performance in terms of both execution time and prediction errors.