A 3-stage fuzzy-decision tree model for traffic signal optimization in urban city via a SDN based VANET architecture

Abstract

One of the main challenges for developed cities is vehicle traffic and its management. Because of physical structures and cross roads of urban intersections, traffic flows may cause time delay, congestion, traffic accidents and more fuel/time consumption. For these reasons, there are many studies on traffic management systems which are a part of smart city applications in both literature and practice. In recent years, it seems that VANET (Vehicular Ad Hoc Networks) architecture has been preferred in these studies, which enables vehicles to easily communicate with each other or devices on the edge of the road and also to transfer the related traffic data to the center. On the other hand, there are many studies on a new network paradigm called SDN (Software Defined Networks) that solves problems such as performance, programmability, scalability, security and management difficulties in today's traditional networks. Depending on all this, in this study, we will present an architectural proposal about how to use SDN and VANET network paradigms together in traffic management systems in order to use both existing intersection management system modules more adaptively and future traffic based services to be added in desired service quality without changing the communication infrastructure. Also in this study, a 3-stage fuzzy-decision tree model was proposed for traffic signaling which is the most important problems of urban traffic by focusing environmental factors (accident situation, priority vehicle transition etc.) and activity information (road construction, meeting, train and bus schedule etc.) not only vehicle density unlike other signaling studies in literature. The proposed model also generates/sends a dynamic flow input to SDN agent vehicles and SDN agent RSU (Road Side Unit) in intersection for making VANET routing protocols (AODV-ad hoc on-demand distance vector, DSDV-destination-sequenced distance vector) selection automatically according to traffic incident and density. In order to measure the functionality of the SDN based VANET architecture and 3 stage fuzzy-decision tree model on it for the transmission infrastructure of the traffic management, performance tests were performed on different traffic and network scenarios on Adapazari City Centre Model. SDN based VANET architecture was compared with traditional systems according to transmission infrastructure and also proposed 3-stage fuzzy-decision model was compared with signaling techiques in literature such as fixed-time signaling, Webster equation, ant colony algorithm and particle swarm optimization under same traffic and network scenarios. The results obtained from performance tests shows that the proposed 3-stage fuzzy-decision model has %15-17 better performance than fixed-time signaling and webster equation and also %9-11 better performance that particle swarm and ant colony optimizations in low density traffic scenarios. In high density and dynamic traffic scenarios these rates reach %20-22 compared to fixed-time signaling and webster equation and %14-15 compared to particle swarm and ant colony optimizations. In addition in this study, proposed SDN based VANET architecture was compared to traditional VANET ITS architectures within the same traffic and network scenarios. According to obtained results, it is shown that the proposed SDN based VANET has %30-40 better performance than traditional systems in end-end delay, throughput and packet loss rate criteria. (C) 2019 Elsevier B.V. All rights reserved.