Distributed System Based on Deep Learning for Vehicular Re-routing and Congestion Avoidance

Abstract

The excessive growth of the population in large cities has created great demands on their transport systems. The congestion generated by public and private transport is the most important cause of air pollution, noise levels and economic losses caused by the time used in transfers, among others. Over the years, various approaches have been developed to alleviate traffic congestion. However, none of these solutions has been very effective. A better approach is to make transportation systems more efficient. To this end, Intelligent Transportation Systems (ITS) are currently being developed. One of the objectives of ITS is to detect congested areas and redirect vehicles away from them. This work proposes a predictive congestion avoidance by re-routing system that uses a mechanism based on Deep Learning that combines real-time and historical data to characterize future traffic conditions. The model uses the information obtained from the previous step to determine the zones with possible congestion and redirects the vehicles that are about to cross them. Alternative routes are generated using the Entropy-Balanced kSP algorithm (EBkSP). The results obtained from simulations in a synthetic scenario have shown that the proposal is capable of reducing the Average Travel Time (ATT) by up to 7%, benefiting a maximum of 56% of the vehicles.