Optimal RSU deployment using complex network analysis for traffic prediction in VANET

Ghosh, Sreya,Chakraborty, Tamal

DOI: https://doi.org/10.1007/s12083-023-01453-5

IF: 3.488

2023-03-09

Peer-to-Peer Networking and Applications

Abstract:Road Side Units (RSUs) are an integral component of Vehicular ad hoc Networks (VANET) along with connected and autonomous vehicles. RSUs have been used to host numerous traffic sensing and control mechanisms to enhance transportation throughput in terms of safety, congestion avoidance, route planning, etc. In order to reduce installation and maintenance costs and associated network and security overhead, it is highly desirable to deploy these RSUs optimally, particularly in strategic and influential positions. While too many RSUs may increase overhead, too few RSUs may fail to map the entire region properly, resulting in erroneous computations. This paper aims to address this trade-off by incorporating a novel scheme called Intersection Influence Analysis System for Optimal RSU Deployment (IIA-ORD). The primary objective of IIA-ORD is achieved through modelling the transportation network as connected graphs and executing a modified K-shell and TOPSIS-based framework. Specifically, the network vertices are mapped with road intersections, and live traffic data is used to analyze various statistical measures, leading to the identification of influential junctions. Extensive performance analysis in an open-source simulation platform backed by real-time data justifies the performance superiority of the IIA-ORD system over existing RSU deployment strategies in terms of an overall number of deployed RSUs, average coverage, coverage time ratio, packet delivery ratio, and delay. The system is validated by a traffic forecasting application. The RSU is equipped with the Stacked Bidirectional Long Short-Term Memory (SBi-LSTM) based traffic prediction model, under which the RSU of a particular junction predicts the traffic congestion of the entire region without the deployment of additional RSUs. Comparative analysis records high accuracy with low loss values for the proposed model in relation to the vanilla LSTM model.

computer science, information systems,telecommunications