SDN-based real-time urban traffic analysis in VANET environment

Abstract

Accurate and real-time traffic flow prediction plays a central role for efficient traffic management. Software Defined Networking (SDN) is one of the key concerns in networking that has caught much attention in recent years. Extensive advancement in software-based configurable equipment has made ready for a new networking paradigm called software-defined vehicular networks (SDVNs). In this paper, we propose a data-driven approach for implementing an artificially intelligent model for vehicular traffic behavior prediction. We combine the flexibility, scalability, and adaptability leveraged by the SDVN architecture along with the machine learning algorithms to model the traffic flow efficiently. First, we introduce an ingenious approach to find congestion sensitive spots in the VANET by means of clustering algorithm and then predicting the future traffic densities for each spot by recurrent neural networks (RNNs). Neural networks (NNs) have been extensively used to model short-term traffic prediction in the past years. We construct a long short-term memory neural network (LSTM-NN) architecture which overcomes the issue of back-propagated error decay through memory blocks for spatiotemporal traffic prediction with high temporal dependency. Due to such learning ability, LSTM can capture the stochastic characteristics and non-linear nature of the traffic flow. The performance evaluation of the proposed approach shows that it has the potential to predict real-time traffic trends accurately with Mean Squared Error (MSE) of 3.0e−3 in the scaled metrics, which is equivalent to 97% accuracy in traffic density prediction.