Computer Networks 185 (2021) 107652 Available online 5 November 2020 1389-1286/© 2020 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Computer Networks journal homepage: www.elsevier.com/locate/comnet A Multi-criteria based handover algorithm for vehicle-to-infrastructure communications Emmanuel Ndashimye a,b,∗ , Nurul I. Sarkar c , Sayan Kumar Ray d a University of Rwanda, College of Science and Technology, Kigali, Rwanda b CMU-AFRICA, Kigali, Rwanda c Department of IT and Software Engineering, Auckland University of Technology Auckland, New Zealand d School of Digital Technologies, Manukau Institute of Technology, Auckland, New Zealand ARTICLE INFO Keywords: Handover V2I communication Multi-tier network Multi-criteria decision making LTE-A Wi-Fi ABSTRACT Maintaining seamless quality of service (QoS) requirements for ongoing delay sensitive applications during handover processes for Vehicle-to-Infrastructure (V2I) communication over heterogeneous networks is a challenging task. The situation becomes even more challenging to deal with for multi-tier heterogeneous networks with various network sizes. To overcome the above challenges, we propose a multicriteria-based handover algorithm for V2I communications (V2I-MHA) to be used in multi-tier heterogeneous network environments. V2I-MHA selects the most appropriate underlying target network for handover based on QoS profiles of the ongoing applications and services. Moreover, our proposed algorithm filters out the inappropriate candidate networks using a multi-criteria decision-making technique based on Simple Additive Weighting (SAW) approach. However, our approach ensures that the most appropriate candidate network for handover processes is selected not only based on QoS requirements of the ongoing applications, but also on the knowledge of candidate network parameters (e.g. bandwidth, packet latency, packet losses, and service pricing). A multi-mode vehicle On-Board Unit (OBU) containing Long Term Evolution-Advanced (LTE- A) and Wi-Fi network interface cards is developed in the Riverbed (previously OPNET) simulator for system performance evaluation. The simulation results show that the proposed V2I-MHA outperforms (in terms of handover failure rate and packet losses) the existing handover methods. 1. Introduction Recently there has been enormous demand for wireless applications and services due to the need for accessing information from anywhere and anytime. This requires wireless communication networks to offer higher data rate [1]. However, the plethora of emerging delay sensitive vehicular network applications, such as, road safety, traffic manage- ment, and infotainment applications, have exacerbated the need for having higher data capacity and bandwidth. Such vehicular network applications require seamless connectivity for moving vehicles. Two of the most common modes of vehicular communications include the vehicle-to-vehicle (V2V) communication in which vehicles connect to each other and exchange information in an ad hoc mode, and the vehicle-to-infrastructure (V2I) communication, in which vehicles connect to fixed roadside infrastructures for communications. With the advent of Fifth Generation (5G) networks [1], layered (multi-tier) network structures having array of cells of different sizes are getting deployed to increase coverage and capacity. In a typical multi-tier network, a wide macro-cell may overlay multiple small cells of different ∗ Correspondence to: KN 67 Street Nyarugenge, Kigali P.O. Box 3900, Rwanda. E-mail addresses: endashim@andrew.cmu.edu (E. Ndashimye), nurul.sarkar@aut.ac.nz (N.I. Sarkar), Sayan.Ray@manukau.ac.nz (S.K. Ray). sizes (femto or pico) to support increased capacity and also to offload network traffic when needed [2]. Despite the different advantages of V2I applications, providing seamless connectivity to moving vehicles remain as one of the main challenges for V2I communications. Currently, the vehicular network infrastructures offer limited coverage, thus preventing in-vehicle users from accessing internet content such as news and weather information while on the move. Standardization efforts in vehicular networks have resulted in Wireless Access for Vehicular Environment (WAVE) protocol. The WAVE protocol integrates the IEEE802.11p at MAC and physical layers; hence, its main limitation is linked to scalability issues [3]. On the other hand, the automotive industry has been build- ing fully featured vehicle On-Board Units (OBU) aiming to integrate different heterogeneous communication technologies such as WiFi, UMTS, LTE, and WiMAX to solve the scalability issue and reinforce the communication systems of vehicles [4]. As shown in Fig. 1, the multi- homing vehicle OBU will have different channels (from different radio https://doi.org/10.1016/j.comnet.2020.107652 Received 13 May 2020; Received in revised form 22 October 2020; Accepted 25 October 2020