Forecasting VANET Traffic Congestion Employing Extreme

Abstract

A vehicular Ad-hoc network (VANETS) is considered as the new age wireless communication that establishes the communication among the vehicles and exchanging data with one another about things like vehicle speed, location & direction. These networks consist of the vehicular onboard unit(OBU) as well as road side unit(RSU) which are used for the exchange of the messages between the vehicles. Though  the VANETS is considered to be an intelligent data exchange between the vehicles,  still it suffers from the non-prediction of traffic congestion which still creates the bottleneck for the creation of an efficient vehicular networks. Deep learning and machine learning algorithms have recently shone a stronger light on studies to improve the network's capacity to predict traffic congestion on the roadways. But these algorithms needs improvisation to be predict the traffic congestion in accordance to the dynamic variation of the vehicles speed and position. Consequently,Extreme Deep Reinforcement Learning techniques(EDRL) is proposed for an effective prediction of traffic congestion under the different scenario of trafficswere used in this work. The traffic scenario was created using the Simulation of Urban Mobility(SUMO) in which datasets for the real-time tests were gatheredand integrated with the proposed model to predict the traffic congestion. Metrics like the average travelling time delay (ATTD) as well as the average waiting time delay (AWTD) are used to assess the performance of the suggested approach. The suggested model's performance is compared to that of various deep learning models already in use, including Convolutional Neural Networks (CNN), Long Short Term Memory (LSTM), and Recurrent Neural Networks (RNN), to demonstrate its superiority .Results demonstrates that the proposed model has shown improved performance over the multiple iterations, since the EDRL is adaptive to the changes in the traffic environment and inclusion of extreme feed forward networks has shown the faster response than the existing deep learning models.