Efficient vehicle-to-infrastructure system for enhancing road safety in fastest-growing cities

Abstract

Vehicle -to -Infrastructure (V2I) are crucial for effective traffic management, offering diverse methods to enhance user awareness of traffic conditions. It is essential to understand the technologies involved and their prerequisites to enable the implementation of V2I technology. The application of the Internet of Things (IoT), and V2I communication and Roadside Units (RSU) are promising paradigms to solve urban transportation challenges. The aim of thesis is to contribute to the advancement of traffic management strategies that employ V2I and their requirements for them to be adopted in the least developed cities. In addition, we propose a predictive model for future road traffic flow rate prediction based on past data for effective traffic intersection management. Various machine learning approaches were used to come up withpromisingmodels. Theobjectiveofthisthesiswereachievedbyapplyingthreeapproaches. Firstly,weconductedasurvey,byinterviewingkeyroleplayersintransportation,questionnaires were distributed and based on individual understanding, perceptions towards the realization of V2I-based technologies, responses were evaluated and tested. The findings showed that the implementation of V2I can reduce traffic congestion and improve traffic flow. Secondly, we employed Machine Learning (ML) and predictive model such as Random forest (RF) and Support vector model (SVM) on the data generated by developed prototype. The models were tested on the collected data by RSU that was deployed to predict the battery current consumption. Thirdly, we employed and compared the performance of two predictive models Long Short-Term Memory (LSTM) and Auto-Regressive Integrated Moving Average (ARIMA) respectively to forecast traffic at one of the selected junction among other busiest junctions using secondary road traffic datasets collected from road Authority in Kigali City. The results revealed that the LTSM is capable of selecting and adjusting the timing of traffic light and could serve as a valuable model to predic traffic resulting in traffic congestion reduction.