Context-Aware Enhanced Application-Specific Handover in 5G V2X Networks

Abstract

The deployment of Augmented Reality (AR) is a necessity as an enabling technology for intelligent transportation systems (ITSs), with the potential to boost the implementation of Vehicle-to-Everything (V2X) networks while improving driver experience and increasing driving safety to fulfill AR functionality requirements. In this regard, V2X networks must maintain a high quality of service AR functionality, which is more challenging because of the nature of 5G V2X networks. Moreover, the execution of diverse traffic requirements with varying degrees of service quality is essential for seamless connectivity, which is accomplished by introducing efficient handover (HO) techniques. However, existing methods are still limited to basic services, including conversional, video streaming, and general traffic services. In this study, a Multiple Criteria Decision-Making (MCDM) technique is envisioned to address the handover issues posed by high-speed vehicles connected to ultra-high-density (UDN) heterogeneous networks. Compared with existing methods, the proposed HO mechanism handles high mobility in dense 5G V2X environments by performing a holistic evaluation of network conditions and addressing connection context requirements while using cutting-edge applications such as AR. The simulation results show a reduction in handover delays, failures, and ping-pong, with 84% prevention of unnecessary handovers.

1. Introduction

The main vision of “Smart Cities” is to maximize the efficiency of urban areas, foster economic expansion, and support the dynamic life requirements of their citizens. The Internet of Things (IoT), which connects objects and exchanges collected data from sensors through the new generation of communication networks, is equipped to improve the availability and reliability of smart city infrastructure. These goals can be achieved by utilizing cutting-edge technology in all service aspects, including health, education, industry, intelligent transportation systems, and community services. Fifth-generation networks have promised the convergence of diverse infrastructures to fulfill the requirements of high data rates, low latency, better coverage, and denser networks. They play a pivotal role in accelerating the implementation of emerging technology that supports advanced smart city applications such as smart manufacturing, connected vehicles, and other life aspects. In addition, the structure of the network architecture based on network virtualization, software definition, and cloud-based computing can facilitate the implementation of critical use cases that provide high data rates and ultra-low latency, allowing the full potential of these cases to be realized. These network features provide a richer communication and sensing interconnection experience for a wide variety of users.

Augmented Reality (AR) is an emerging technology that benefits from these features and unlocks their innovations. It is a technique that enhances the reality of an object in its natural context by linking it to digital technology and providing users with the ability to sense reality through computer-generated visuals. It is a subset of the Mixed Reality (MR) continuum, where Virtual Reality replaces the physical world, whereas, in contrast, AR enhances the effect of realism by importing and integrating virtual objects into the real environment [1]. Immersive technologies have led to advancements in the automotive industry. AR applications are steadily being injected into ITS to enhance the driving experience by assisting drivers with navigation and displaying road and vehicle real-time information to improve safety concerns.

1.1. The Integration of AR Applications for Vehicle-to-Everything (V2X)

The increased integration of vehicles into human lives, not only as a means of transportation, has resulted in an increase in safety issues under both road and driving conditions. Mobile networks have evolved to support broadband communications that require high data rates, improved network coverage, high reliability, and ultra-low latency. Further improvements in network operations that reduce management complexity and enable reliable handover operations and seamless network transfers can result in the integration of multimedia applications in automotive use cases that require speed, responsiveness, and reliability for data transfers.

V2X is a type of communication between transport vehicles (V2V), pedestrians (V2P), infrastructure (V2I), and networks (V2N) that greatly benefit from the features of 5G networks, particularly those that involve the implementation of applications focusing on road and traffic safety. AR is one such application that is expected to improve V2X safety using digital information and objects that are overlaid on roads or traffic environments [2]. The deployment of AR in V2X use cases, as presented in Figure 1, attributed to 5G Advanced (3GPP Release 18, building on Releases 15, 16, 17), will bring notable developments that focus on the following topics [3,4]:

▪

Intelligent network automation: This network provides AI/ML implementation for intelligent network operations.

▪

Extended reality (XR): This concept refers to future immersive technology that enhances the capabilities of MR and its technologies by allowing for multi-sensory interactions. It aims to increase human sensory perception by augmenting the existing world with additional information or by constructing entirely new environments. It benefits from high-data-rate communication and Low Latency Low Loss (L4S), which allows latency to be prioritized over data rates in the event of traffic congestion.

▪

Reduced Capability (RedCap) NR Devices: RedCap devices focus on Broadband IoT use cases, allowing reduced modem costs, design relaxation, leaner procedures, and support for extended discontinuous reception (eDRX).

▪

Network Energy Savings: This concept focuses on traffic load balancing and sleep mode for Next-Generation Node B (gNB) to enhance network energy-saving schemes.

▪

Deterministic Networking for IoT: This network provides support for use cases where there are heavy requirements for bounded low latency, low-delay variation, and extremely low losses.

The integration of AR applications in V2X will become more prominent in the following areas, as enabled by the features of 5G networks [5]:

▪

Driving Assistance and Safety: AR’s seamless access to information storage enables the real-time rendering of visual cues in the road environment, enhancing awareness and attention through alerts for emergency behaviors, bad road conditions, and unfavorable weather patterns.

▪

Navigation: AR can be used to access real-time information to show navigation aids that not only provide proper guidance but also reduce unwanted distractions.

▪

Human–Machine Interface: AR fosters the development and use of output devices such as windshield displays (WSDs) and heads-up displays (HUDs) that enhance the visual fidelity of the driving environment. AR can also provide information such as driving conditions, behavior, and intent externally to vulnerable road users such as pedestrians and cyclists.

▪

Passenger Experience: AR can provide in-vehicle passenger instructions and entertainment that improve comfort and ease anxiety.

1.2. AR Implementation Constraints for V2X

To effectively integrate AR into V2X systems, some of the key challenges related to infrastructure and interoperability must be addressed. Handover (HO) is a process that is central to the maintenance of seamless and uninterrupted communications, especially in 5G V2X networks where UEs move rapidly across the network. HO ensures that UEs are continually connected with the best signal quality and low latency. These connection qualities are critical to vehicular applications such as autonomous driving and Augmented Reality (AR). The HO process must be effective enough to reduce packet loss, delays, and connectivity failures to enhance system reliability and user experiences in V2X environments. The implementation of heterogeneous networks (HetNets) creates complexities that could hinder the efficient implementation of V2X technologies in all cases of use. HetNets comprise diverse types of dense small cells (SCs) within a short range of areas, resulting in the frequent switching of connections and link failure. Owing to high network density, traditional handover techniques are found to be inadequate and/or inefficient in V2X networks, where User Equipment (UE) generally moves at high speed, and HO failure rates and service interruptions, ping-pong, and unnecessary handovers increase exponentially [4,5]. The ping-pong occurs when the UE connects to a new cell and then reverts back to the source cell [6]. Network congestion causes handover inefficiencies that affect resource allocation. Quality of service (QoS) is a requirement that needs to be addressed, and traditional HO usually does not address QoS. Latency requirements are highly significant in AR V2X networks, and traditional HO can result in a higher delay, which can negatively impact safety and low-latency requirements; the seamless handover in HetNets is challenging in these aspects. A predictive and proactive handover is now a requirement for a high-speed network environment. A mechanism that allows handovers to scale well is required to optimize the connectivity and maintain a reliable connection for AR applications in a V2X environment.

1.3. Problem Definition and Main Contributions

The use of 5G V2X networks promises to address the high mobility demands of UE, especially in terms of seamless handover performance. However, this promise comes with a number of complexities, and despite advancements in architecture and resource allocation, mobility management presents notable issues, particularly in terms of UE application usage. Augmented Reality (AR) has been touted as one of the potential solutions that can boost driving experience and safety. Using this technology, however, increases the complexities of high mobility that largely impact the handover processes, leading to increased latency and unnecessary handovers, as well as failures and ping-pong effects. Due to the limitations of traditional HO that have been significantly highlighted by recent studies, HO algorithms must be improved or enhanced to adapt to dynamic network conditions in high-speed V2X networks. Current HO techniques rely profoundly on signal strength-based measurements, which are proven to be insufficient in addressing the varied QoS requirements of vehicular applications such as AR and autonomous driving. High-speed UEs trigger frequent and unnecessary Hos, which is typical in dense 5G networks with a high number of SC deployments, leading to performance degradation. These issues significantly call for an intelligent HO solution that leverages the use of real-time user-specific context information in a predictive modeling approach to alleviate negative effects.

The gaps that have been presented highlight the need for a novel method that utilizes rich contextual information from UEs, specifically application data, that intelligently inform handover decisions. An enhanced handover mechanism that integrates application contexts such as AR and network parameters such as signal metrics, UE speed, and SC load into a Multiple Criteria Decision-Making (MCDM) model will result in a more holistic and informed HO decision.

The main contributions of this paper are discussed as follows:

▪

A user context-aware process is proposed that considers the service requirements (connection context) of UEs in a dynamic, high-density network environment to optimize handovers while considering basic services such as conversation, video streaming, general traffic (browsing), offloading data, relay services, and advanced technologies such as AR.

▪

The appropriate assignment of weight-to-criteria parameters in MCDM remains an open issue [7]. It is recommended that the weighting should be dynamically assigned in accordance with the network topology changes. In our proposed MCDM-based HO procedure, we consider the dynamic assignment of weights to an individual network connection quality metric based on network type, and the connection context policy is provided.

▪

The enhancement of network performance is addressed by reducing the number of handover issues for specific V2X applications. To the best of our knowledge, there is currently no existing study that examines the handover issues for 5G V2X while using AR applications.

▪

The simulation findings indicate that the proposed method maintains consistent performance across different UE speeds, owing to the effective selection of cells that are adaptable to the diverse needs of the network QoS compared to the recent study in [8].

The remainder of this paper is organized as follows. Section 2 discusses related work and highlights the solutions and improvements. Section 3 and Section 4 discuss the proposed algorithm and findings, respectively. Finally, the conclusions are presented in Section 5.

2. Related Work

Recent research papers have examined the role of 5G networks in facilitating the deployment of V2X technologies, addressing different challenges, proposed solutions, and future work [9,10]. Handover issues are one of the key challenges that significantly affect the reliability and continuity of V2X communications. To extend the network life and improve its reliability, where the UE has high mobility, a more sophisticated HO algorithm needs to be deployed to detect the context of the network required and the type of UEs and use cases requiring network connection [11]. Responsive HO for hybrid mobility management requires context-aware mechanisms. This context can be used to tailor the fit of resource allocation and management dynamically to the type of UE or V2X application requirements. The UE connection context is crucial for cell selection in V2X to optimize the handover for the following reasons:

▪

Devices in a V2X network have mobility patterns that exhibit varying speeds. In addition to knowing the direction and speed of UEs, the connection context helps efficient handover decisions by adjusting the connectivity requirements.

▪

The connection context embodies the required network conditions, data requirements, latency, and reliability that are important to ensure that a handover is performed without interruption to a relevant base station or SC.

▪

The connection context allows handovers to be optimized by anticipating network loads and performance qualities.

▪

The connection context establishes requirements for service quality and the prioritization of applications, ensuring that the handover of UEs is made to highly capable base stations or SC.

Generally, it is considered that the connection context in handover is important in guaranteeing that the handover is relevant, reliable, and efficient while catering to the exclusive requirements of UEs with better mobility and user satisfaction. MCDM is considered an effective approach for modeling HO decisions in HetNets, resulting in a reduction in HO problems caused by an improper small-cell selection method [12]. The different MCDM solutions, such as the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), Simple Additive Weightage (SAW) [13], and Analytic Hierarchy Process (AHP) [14], are widely implemented to select the most suitable SC for the handover procedure. However, the practice of assigning a fixed weight to each HO parameter in the MCDM method causes the UE to make an inappropriate decision for cell selection in terms that do not reflect the changes in the user resident networks. This, in turn, leads to an increase in the number of handovers, which can add more signaling load to a network [12]. MCDM weights can be classified into three categories: subjective, which relies on expert preference or opinions; objective, which uses mathematical methods instead of expert judgment to determine the relative importance of criteria; and hybrid weighting [15]. In [16], the MCDM was followed to form a context-aware mechanism. AHP, a pairwise comparison approach, is applied to determine the criterion subjective weight based on the UE preference. By utilizing TOPSIS, network criteria and user performance are both involved in selecting the best Radio Access Technology (RAT). Network criteria and user performance are both involved in selecting the best RAT for the connection. Overall, the mechanism eliminates unnecessary handovers by reducing the frequency of handovers. In [17], the MCDM approach is also followed for handover optimization in 5G dense networks. They proposed an enhanced Multi-Objective Optimization Method (MOOR) and a Q-learning method for a two-tier HetNet architecture. To avoid abnormality ranking in the MCDM, the ratio analysis approach (MOORA) merged with the enhanced Entropy objective weighing method and vector normalization instead of summation normalization. The Q-learning method is used to incorporate the most efficient triggering times during the user movement. The computational complexity of this approach introduces additional delays that consequently slow down the handover decision, which is critical for high-mobility users. To optimize the HO decision, a hybrid MCDM could be a more effective way to set the parameter weight values for the cell selection method. Hybrid MCDM is a combination of subjective and objective weighting approaches. A novel opportune context-aware network selection algorithm (OCAN) is proposed in [18] to maximize user satisfaction for the selection of the best network. The handover strategy is developed based on the user satisfaction indicator (USI) and the stability of the best network. The USI evaluates the potential networks by considering the diversity of context parameters, including user preference, energy state, user mobility, and conventional application-specific parameters. This approach is employed using the Analytic Hierarchy Process (AHP) and utility function theory. It effectively combines user and network factors. The relevant parameters are traffic organized in hierarchal multi-layer structures, where the USI is at the top layer, serving as a global indicator that is connected to a number of attributes. The author in [19] proposed a hybrid MCDM method by combining the Fuzzy Analytical Hierarchy Process (FAHP) to acquire the weight of the handover metrics and the Fuzzy Technique for Order Performance by Similarity to the Ideal Solution (FTOPSIS) method to rank cells as determinants for the best HO target. The priority level for different traffic classes is considered with the aim of reducing the number of unnecessary HOs using closer coefficient values for the current and best networks. However, this approach is implemented for conventional dense networks and does not consider high-mobility users (V2X). Despite the fact that studies have employed various methods to assign weight to criteria, the assigned weight stays constant across different networks.

Table 1. A comparative summary of MCDM-based handover decisions.

Criteria	Pairwise Comparison [16]	E-MOORA [17]	OCANS [18]	Fuzzy-Based MCDM [19]	CAP (Proposed Method)
Weight method	Subjective	Objective	Hybrid	Hybrid	Dynamic/adaptive
Weight assignment	AHP	Modified Entropy	AHP and utility function	FAHP	Context-connection policy-based method
Cell selection	TOPSIS	Q-Learning	USI metrics	FTOPSIS	Performance index, stay time, load
Adaptability to network changes	Yes (based on UE priority)	Yes	Limited	No	Yes
Computation complexity	Moderate	High	High	High	Low
Latency	Moderate	High	High	High	Low
Heterogeneity	Yes	Yes	Yes	Yes	Yes
UDN	Yes	Yes	Yes	No	Yes
Limitations	Criterion weight based on decision-makingpreference	Computation overhead in dense networks results in more delays	Computation overhead in dense network results in more delays	Criterion weight based on making-decision preference and computation overhead	Optimizing handover decision using ML

provides a comparative summary of MCDM-based handover decisions.

In [8], a technique is presented that integrates adaptive context-aware algorithms, efficient scanning, load balancing, and hybrid channel allocation to improve vertical handover decisions in heterogeneous networks. The fundamental feature of the proposed mechanism is its flexible response to changing network conditions and user connectivity requirements. The mechanism uses static network elements such as network topology and dynamic or continuously altered features such as network traffic and mobile node location data to optimize handover decisions, thereby lowering the possibility of missed calls and delays. A network scanning approach chooses when and what to scan depending on the possibility that a handover is required soon instead of continuously searching through available networks. This lessens the energy consumption of mobile nodes. A utility function evaluates the load of network cells, allowing resources to be used more effectively and avoiding bottlenecks that are caused by cells equally distributing the load. Finally, the hybrid channel allocation strategy prioritizes handover traffic by dynamically allocating channels according to the priority of the traffic type, ensuring that vital communications continue uninterrupted throughout the handover process, which is especially crucial in heterogeneous networks because various types of traffic may need different levels of quality of service. In [20], the authors utilized reinforcement learning (RL), based on Q-learning, to optimize handover decisions. Overall, this method improves QoS by minimizing handover failures, reducing ping-pong handovers, and increasing network throughput. Nevertheless, this intelligent algorithm must maintain sufficient iterations to achieve optimal handover decisions, thereby increasing the likelihood of time processing [21] and computation complexity, which is undesirable for high-mobility users such as the V2X network.

Recent studies have focused on examining the optimality of cell selection in wireless networks and have suggested different mechanisms with which to tackle the performance degradation caused by handover issues. As listed in

Table 2. A comparative summary of handover decision strategies.

Ref.	Network Selection Strategies					Related Parameters			HO KPIs	Handover Decision Criteria
	Single Attribute	Utility Theory	Multi-Criteria	Markov Model	Intelligent Based	Network	User Terminal	Traffic Type
[16] (2017)	×	×	√	×	×	RSS, data rate, Jitter, packet loss, Jitter delay	×	Basic traffic	Average number of handovers	Highest-ranking cell
[17] (2021)	×	×	√	×	√	SINR, Delay, load, transmitted power	User movement	×	Average number of handover and link failures	Highest-ranking cell
[18] (2022)	×	√	√	×	×	RSS, bandwidth, security grade, packet loss, delay, Jitter	user velocity, distance, Energy, cost	Basic traffic	Average number of handovers	Maximum USI and successive time (stability of the best network)
[19] (2023)	×	×	√	×	×	RSSI, SINR, delay, packet loss, BER	Energy, speed, direction, working mode	Basic traffic	Average number of handovers	Fuzzy combined decision matrix (closer coefficient threshold)
[8] (2023)	×	√	×	×	×	RSS, SINR, BER, call arriver, bandwidth	Speed, location, straight trajectory	Basic traffic	Average number of handovers delay and link failures	Dwell time, availability of allocated channel
[20] (2024)	×	×	×	×	√	RSRP, RSRQ, SINR	UE Speed	Network slicing	HO ping-pong, radio limk failure and throughput drop	Histrorical Q-Learning values
Proposed method	×	×	√	×	×	RSSI, SINR, BER, data rate, delay, packet loss	Velocity, location, dynamic user movement	Basic service, AR, data offloading, relay traffic	Average number of handovers, HO successful rate, radio link failure, delay, ping-pong HO average rate, and necessary and unnecessary HO average rate	TOPSIS performance index, stay time, load

, none of these solutions have specifically introduced the handover challenges faced by V2X users in dense networks while using AR technology. This prompted us to design a context-aware system that considers instances where the network topology frequently changes while using emerging applications such as AR to lessen the probability of handover issues in UDN-V2X networks. In this study, MCDM weighting was dynamically adjusted in accordance with the network topology changes, leveraging policy context connection and aiming for efficient handover decision-making compliant with the performance requirements of 5G V2X networks.

3. Connection-Aware Policy Mechanism (CAP)

In 5G V2X networks, it is critical to guarantee uninterrupted connectivity. Vehicles always move and interact with different network nodes; therefore, an effective handover mechanism is essential. Conventional handover methods frequently fall short of the dynamic demands of contemporary V2X applications. In this scenario, the TOPSIS-Based Connection Context-Aware Policy Mechanism (CAP) is utilized, providing an advanced method to improve handover through the utilization of several network connection quality parameters and the dynamic modification of their weights according to the connection context. The mechanism is intended to be deployed using the Edge Computing paradigm, which is a distributed approach that localizes computation and data management. Implementing this type of computing model allows the mechanism to be deployed near the location where it is most needed (end-user devices) to process data closer to where it originates. This results in reduced latency, optimized data processing and transmission, reliable decision-making, scalable handling of data loads, connection-context awareness, and lower power consumption. The Connection-Aware Policy (CAP) algorithm is presented below (Algorithm 1), followed by Figure 2, which illustrates the conceptual visualization of the CAP mechanism.

Algorithm 1: The Connection-Aware Policy Mechanism (CAP) Algorithm
1	Model the network grid
2	Identify UEs and plot their movements in network grid sector
3	Input:
	- List of SCs in UE Sector ID No (m, n) = ([x/20], [y/20])
	- User Connection Context (e.g., AR, VIDEO, AUDIO)
4	Output: List of Filtered SCs
5	Set initial parameter weights (W) based on UE context
	Example:
	If context == “AR”       → W = [RSSI = 0.1, SINR = 0.1, BER = 0.1, Delay = 0.3, PLR = 0.2, DataRate = 0.2]
	If context == “VIDEO”   → W = [RSSI = 0.2, SINR = 0.2, BER = 0.1, Delay = 0.1, PLR = 0.2, DataRate = 0.3]
	If context == “AUDIO”   → W = [RSSI = 0.3, SINR = 0.3, BER = 0.1, Delay = 0.2, PLR = 0.1, DataRate = 0.0]
6	while List of SCs[i] <= List of SCs[max] do
7	Extract parameter (RSSI, SINR, BER, Delay, PLR, DataRate) values from SCs[i]
8	Construct decision matrix X
9	Store parameter values of SCs[i] and normalize the weights in Xij
10	Formulate the decision matrix X = X−ij * Wj using dynamically adjusted weights W
11	Calculate Performance Index (PI) Pi = √(Σ Xij2), j = 1,...,6
12	end while
13	Rank SCs based on Performance Index Rn = arg max Pi(i)
14	Filter top n percentage of SCs (20%) Top⌈0.2 × n⌉(Cranked)
15	Calculate Stay Time Assigned STA = d/v
16	Calculate Stay Time (ST) for SCs STi = STA − (Ci + Ri)
17	for all SCs in Top⌈0.2 × n⌉(Cranked)
18	while HO required = true and Current SC = None
19	if Detected Context ≠ Previous Context then
20	Update W
21	Recompute Xᵢⱼ and PI for Current SCs using updated W
22	Query SC Load
23	if SC Load < Max SC Load then
24	if ST <= STA then
25	Current SC = SC
26	Increment SC Load by 1
27	Decrement ST by 1
28	else
29	Current SC = None
30	end
31	end
32	end while
33	end for

3.1. Network Connection Quality Parameters

The following crucial parameters are used by the proposed mechanism to evaluate the SC performance:

▪

Received Signal Strength Indicator (RSSI): This device calculates the power of the radio signal ratio that is received.

▪

Signal to Interference and Noise Ratio (SINR): The signal quality is measured by comparing the strength of the desired signal to noise and undesired interference.

▪

Bit Error Rate (BER): This indicates the number of incorrect bits received during the transmission.

▪

Data Transmission Rate: This represents the rate at which information is sent.

▪

Delay: The amount of time it takes for data to move from one location to another is calculated.

▪

Packet Loss: This represents the quantity of packets dropped during the transfer.

▪

Each of these factors is essential in determining the overall performance of an SC.

3.2. Application-Specific Connection Context Policy

The connection context policy dynamically modifies the weight of network connection quality characteristics in accordance with predetermined performance standards. The priority decision for basic services has been defined by the 3GPP standard [22] and evaluated by rigorous study of the recent literature according to the given references [13,19,23] and [22,24,25,26] for recommended traffic classification. This policy specifies a range of codes that meet the demands of diverse applications as follows:

▪

Audio: This guarantees uninterrupted and clear communication, and it is optimized for audio services.

▪

Video: This adjusts the quality of the connection for uploading and streaming videos.

▪

General: A well-balanced quality for standard network utilization is offered.

▪

AR: AR applications are needed to necessitate the delivery of excellent graphics and videos.

▪

DataOff (Downloading and Uploading Data): This gives connection quality the first priority for services that require a lot of data, including uploading and downloading.

▪

Relay: This improves services in which a UE serves as a network relay.

The UE executes a set of traffic profiles based on its requirements.

Table 3. Requirements of connection context policies.

User Connection Traffic Class	Important Decision Criteria
Audio	Minimal Latency: To guarantee real-time audio connection, minimal latency is prioritized. Moderate SINR and Packet Loss: Signal deterioration and packet loss are tolerable for audio quality.
Video	High Data Throughput: Streaming video calls for high transmission speeds. Low to Moderate Latency: A small amount of latency is acceptable for video streaming. Minimal Packet Loss: Guarantees seamless video playback
General	Balanced Parameters: To provide universal network connectivity, a balanced set of parameters is desired.
AR	High Data Transmission Rate: To broadcast high-definition images and videos, high throughput is needed. Low Latency: To keep AR apps responsive, there must be little delay. Low Packet Loss: This preserves the visual content’s integrity.
DataOff	High Data Throughput: Gives large-scale data uploads and downloads priority over the data transmission rate. Moderate Delay Tolerance: When the throughput is large, a certain amount of delay is acceptable.
RELAY	High SINR and RSSI: Gives priority to clear, powerful signals in order to guarantee dependable relay operation. Moderate Packet Loss and Delay: Relay functions are tolerant of a small amount of packet loss and delay.

illustrates the perceived requirements of connection context policies.

Furthermore, the weight adjustments are contingent upon the network environment type (urban, suburban, or open) and represent the distinct challenges and requirements associated with each setting.

(a)

Urban Networks:

▪

High User Density: Due to the high user density in urban locations, there may be interference and congestion.

▪

Building Obstructions: Signal deterioration and multipath propagation can be brought on by buildings and other structures.

(b)

Suburban Networks:

▪

Moderate User Density: Compared with metropolitan regions, suburban areas have a lower user density, which reduces traffic.

▪

Fewer Obstructions: Signal quality is improved because there are fewer obstructions than in metropolitan areas.

(c)

Open Areas:

▪

Low User Density: The least densely populated locations are those found in open spaces, such as rural or freeway settings.

▪

Minimal impediments: Better signal transmission is often achieved in these locations because of the lack of major impediments.

3.3. Small-Cell Performance Index (PI) Measurement Using TOPSIS

The TOPSIS criterion-weighted result is deemed to be more accurate compared to other MCDMs, demonstrating a high level of accuracy [27] due to the rationality of human decision-making [28]. This study offers a context-aware method as an enhanced version of our previous study [29]. The MCDM-based mechanism aims to select the best target SC to maximize HO success, minimize disruptions, and ensure optimum user experience using a set of heterogeneous criteria that are relevant to a UE connection context. Each criterion is classified based on whether it degrades or improves connection quality. The maximization of criteria such as RSSI, SINR, and the data rate is aimed at improving connection capacity or strength. Minimization focuses on the reduction in latency and loss (BER, delay, packet loss rate, and load). Values are normalized [0,1] for the benefit of the criteria. Higher values are beneficial for RSSI and SINR, while inversed values are beneficial for BER and delay, for example. This ensures that all scores align with a maximization objective. Once the values are normalized, they are aggregated into the PI value through a weighted sum model. This decision-based framework integrates goal setting, criterion classification, normalization, and scalar aggregation to ensure that the MCDM process is clear and robust enough to indicate that the mechanism is not based on weights alone. Dynamic-weighted network connection quality metrics are used to calculate PI for every SC using the TOPSIS method. To provide a clear rating of SCs from best to worst, TOPSIS finds the SC that is closest to the ideal solution and farthest from the negative ideal solution as per the following steps:

• Define the Network Connection Quality Parameters

Let X be the decision matrix where X_ij is the value of the parameter P_j for SC, and the network parameters are denoted as follows: P₁: RSSI, P₂: SINR, P₃: BER, P₄: data transmission rate, P₅: delay, and P₆: packet loss.

• The normalization of weights is calculated as follows:

$$X_{ij}^{\prime }=\{\begin{array}{cc}{\displaystyle \frac{X_{ij}-min\left(X_j\right)}{max\left(X_j\right)-min\left(X_j\right)}}& \mathrm{if} j \mathrm{is} \mathrm{to} \mathrm{be} \mathrm{maximized}\\ {\displaystyle \frac{max\left(X_j\right)-X_{ij}}{max\left(X_j\right)-min\left(X_j\right)}}& \mathrm{if} j \mathrm{is} \mathrm{to} \mathrm{be} \mathrm{minimized}\end{array}$$

(1)

• Then, the dynamically adjusted weights are determined:

$$\mathrm{Let} W_{ijk}={\alpha }_{ij}.{\beta }_{ik}$$

(2)

Here, α_ij is the base weight for parameter P_j in network type i, and β_ik is the adjustment factor for parameter P_j based on the connection context k.

• Then, the decision matrix is formulated:

$$\mathrm{X}=\left[\begin{array}{cccccc}{x}_{11}& x_{12}& x_{13}& x_{14}& x_{15}& x_{16}\\ x_{21}& x_{22}& x_{23}& x_{24}& x_{25}& x_{26}\\ x_{31}& x_{32}& x_{33}& x_{34}& x_{35}& x_{36}\\ x_{41}& x_{42}& x_{43}& x_{44}& x_{45}& x_{46}\\ \cdots & \cdots & \cdots & \cdots & \cdots & \cdots \\ x_{\mathrm{n}1}& x_{\mathrm{n}2}& x_{\mathrm{n}3}& x_{\mathrm{n}4}& x_{\mathrm{n}5}& x_{\mathrm{n}6}\end{array}\right]$$

• The Ranking and Stay Time Assignment (STA) is as follows:

After evaluating the quality of the network using TOPSIS, the ranked list of SCs is filtered by determining the top n%, which lowers the density and ensures that only the highest-performing cells are selected. This step is essential for effectively managing network resources and preventing congestion. Top_k represents the function used to obtain the top k elements from a sorted list. For example, the equation to obtain the top 20% of SCs is as follows:

$${\mathrm{Top}}_{[0.2\times \mathrm{n}]}\left(\mathrm{Cranked}\right)$$

(3)

where C_ranked is the list of PI values sorted in descending order.

A stay time (ST) value is assigned to the selected SCs, which represents the ideal amount of time needed to maintain the connection with the selected SC before allowing a handover to a new SC. As the UE moves and conditions change, this ST allotment aids in preserving a strong, reliable connection. This approach significantly improves the QoS in the UDN network by reducing needless handovers, signaling overhead, and potential changeover failures.

The following is given:

▪

v: velocity of the UE;

▪

d: distance between previous and current UE movements;

▪

C_i: PI of SC_i;

▪

R_i: rank of SC_i.

• The Stay Time Assigned (STA) for the network sector is calculated as follows:

$$\mathrm{STA}={\displaystyle \frac{d}{v}}$$

(4)

2.

The ST for each SC_i in the network sector is calculated as follows:

$${ST}_i=STA-(C_i+R_i)$$

(5)

The handover process’s initiation depends on achieving the handover decision criteria, which consists of the highest PI, the assigned ST, and the availability of resources. The top-ranked SC is subject to load checking. The load ratio is based on the (available connection/total connection). Based on this, the ST threshold is checked, and subsequently, the handover process is initiated.

3.4. Justifications for the Dynamic Adjustments of the Parameter Weights

The varied and changing needs of contemporary V2X applications justify the dynamic adjustments of the network connection quality parameter weights in the TOPSIS process. Conventional mechanisms following static weight allocations are unable to meet the diverse requirements of various environments and services. For example, if the network is built for low latency, which is crucial for AR applications, then an application demanding high data rates, such as video streaming, will not run optimally. The suggested mechanism ensures that the most important criteria are given priority by dynamically modifying the weights, which significantly increases the quality of services.

Benefits

▪

Improved service quality: This technique guarantees the best possible performance for a variety of services by customizing the connection parameters to match certain applications.

▪

Adaptability: The system’s dynamic adjustment enables it to change according to the needs of applications and the network environment.

▪

Efficiency: The most effective use of network resources is ensured by reducing the density of SCs to the highest-performing ones.

▪

Better handover: The TOPSIS rating system and the ST value provide fast and seamless handovers, lowering the possibility of lost connections and enhancing the overall user experience.

Tradeoffs

▪

Complexity: System computation is complicated by dynamic adjustment and TOPSIS computations.

▪

Operational overhead: The network management procedure may incur additional operational overhead due to ongoing weight adjustments and monitoring.

▪

Initial setup: Defining and adjusting the connection context policies for various environments and applications can take a lot of time and testing.

The following mathematical formula handles the assignment of dynamic weights to each network connection quality metric based on the type of network (open, suburban, urban) and the type of connection context policy (Audio, Video, General, AR, DataOff, Relay).

The network connection quality parameters are denoted as follows: P₁, P₂, P₃, P₄, P₅, P₆.

W_ijk represents the weight assigned to parameter P_i for network type _j and connection context _{k. i} ranges from 1 to 6 (representing the count of the network connection quality parameters); _j ranges from 1 to 3 (representing the network types); and _k ranges from 1 to 6 (for the six connection context policy codes).

The equation for the weight assignment is Formula (1) (${\mathit{W}}_{\mathit{i}\mathit{j}\mathit{k}}={\alpha }_{\mathit{i}\mathit{j}}.{\beta }_{\mathit{i}\mathit{k}}$)

Here, the following definitions are given:

▪

${\alpha }_{\mathit{i}\mathit{j}}$—presents the base weight of the parameter P_i in network type _j.

▪

${\beta }_{\mathit{i}\mathit{k}}$—presents the adjustment factor for parameter P_i based on connection context k.

Base Weights (α_ij)

The significance of each parameter in the various network types is represented by these weights. To keep things simple, we can create a base weight matrix A with specified elements αij based on the following network conditions:

$$\mathrm{A}=\left[\begin{array}{ccc}{\alpha }_{11}& {\alpha }_{12}& {\alpha }_{13}\\ {\alpha }_{21}& {\alpha }_{22}& {\alpha }_{23}\\ {\alpha }_{31}& {\alpha }_{32}& {\alpha }_{33}\\ {\alpha }_{41}& {\alpha }_{42}& {\alpha }_{43}\\ {\alpha }_{51}& {\alpha }_{52}& {\alpha }_{53}\\ {\alpha }_{61}& {\alpha }_{62}& {\alpha }_{63}\end{array}\right]$$

Example

▪

In urban networks (j = 1), SINR and delay might be more critical due to higher interference and congestion rates.

▪

In suburban networks (j = 2), the RSSI and data transmission rate might be more important.

▪

In open areas (j = 3), BER and packet loss might be prioritized.

Adjustment Factors (β_ik)

These variables modify the weight of each parameter based on the connection context policy. An adjustment factor matrix B was created with each element βik specified according to the needs of the application:

$$B=\left[\begin{array}{cccccc}{\beta }_{11}& {\beta }_{12}& {\beta }_{13}& {\beta }_{14}& {\beta }_{15}& {\beta }_{16}\\ {\beta }_{21}& {\beta }_{22}& {\beta }_{23}& {\beta }_{24}& {\beta }_{25}& {\beta }_{26}\\ {\beta }_{31}& {\beta }_{32}& {\beta }_{33}& {\beta }_{34}& {\beta }_{35}& {\beta }_{36}\\ {\beta }_{41}& {\beta }_{42}& {\beta }_{43}& {\beta }_{44}& {\beta }_{45}& {\beta }_{46}\\ {\beta }_{51}& {\beta }_{52}& {\beta }_{53}& {\beta }_{54}& {\beta }_{55}& {\beta }_{56}\\ {\beta }_{61}& {\beta }_{62}& {\beta }_{63}& {\beta }_{64}& {\beta }_{65}& {\beta }_{66}\end{array}\right]$$

Example

▪

For Audio (k = 1), SINR and delay might be prioritized

▪

For AR (k = 4), the data transmission rate and delay are crucial.

▪

For DataOff (k = 5), the data transmission rate and delay might have higher factors.

Calculation Example:

To determine the weight for each parameter P_i in an urban network (j = 1), the connection context is AR (k = 4).

• Find the base weights for urban networks:

$${\alpha }_{11},{\alpha }_{21},{\alpha }_{31},{\alpha }_{41},{\alpha }_{51},{\alpha }_{61}$$

2.

Find the adjustment factors for the AR context:

$${\beta }_{14},{\beta }_{24},{\beta }_{34},{\beta }_{44},{\beta }_{54},{\beta }_{64}$$

3.

Compute the final weights:

$${\mathit{w}}_{\mathit{i}14}={\mathit{\alpha }}_{\mathit{i}1}\cdot {\mathit{\beta }}_{\mathit{i}4} \mathrm{for} \mathit{i}=\mathrm{1,2},\mathrm{3,4},\mathrm{5,6}$$

This formula can be used to dynamically assign weights to network connection quality parameters based on the connection context policy and type of network being used. By adapting the weights to certain circumstances, the proposed mechanism can significantly increase the reliability and performance of handovers in 5G V2X networks.

4. Results and Discussion

The 5G-UDN V2X environment simulation was developed using Python 3.10.13 to test and evaluate the effectiveness of the CAP method in selecting the best candidates. It offers a comparison of the probability of similar HO issues between the proposed mechanism and the other two algorithms: the traditional handover and ACAVHD [8].

Table 4. The setting value of the experiments.

Parameters	Values/Ranges
Size of the area	(100 × 100) km2
Network sectors	25
No. of SCs in each sector	30, 50, 100
No. of UEs	10, 20 users per sector
UE speed	20~100 km/h
Connection quality parameters	- RSSI range for 5G NR (–120 dBm to −13 dBm) - SINR range (0–127dB) - BER 10-9, 10–13 ratio - Data transmission rate (100–400 Mbps; mid-band 5G)
Connection quality Parameter weights	Assigned randomly; normalized
Maximum distance	19 km
Data rate	100–400 Mbps; mid-band 5G
V2X 5G E2E latency	7.8 ms
User connection traffic classes	Audio, Video, General, AR, DataOff, Relay transmission.

lists the network configuration settings that are considered in the testing procedure. The behavior of vehicle movement is represented by the Random Walk mobility model [30]. The vehicle traverses the network area by randomly changing its direction and velocity at distance intervals. SCs that are not in the direction of the vehicle movement or have a low PI are filtered out and excluded from the final decision handover process. Typical 5G NR configuration parameters are assigned randomly to SCs. Mid-band 5G bands are utilized to support the applications of intelligent transportation systems [31]. The MEC-based end-to-end V2N2V is the modeling component considered for V2X 5G latency [25].

The simulation was developed using Python 3.10.13 and the Numpy, Matplotlib, Pandas, and OpenPyXL libraries. The experiments were conducted in a simulated 100 × 100 km² network grid, which was subdivided into 25 network sectors. The number of UEs was set at 10 and 20 per sector, respectively, with varying numbers of SCs. The performance of SCs was evaluated by extracting QoS metrics (RSSI, SINR, BER, data rate, delay, and packet loss) had their weights assigned according to the UEs application context (Audio, Video, General, AR, DataOff, Relay). These contexts serve as the decision-making policy used in the TOPSIS process.

Each testing scenario allows for the adjustment of the UE numbers, UE movements, and the number of SCs within a sector. The UE mobility functionality considers the UE’s ability to move at different angles between the sectors. This displays an imitation of a real V2X environment.

Figure 3 shows the application-specific dynamic weight adjustment of each parameter examined for the mechanism testing scenario in accordance with UE traffic communication requirements, ensuring that the total for each connection is equal to one.

The performance of the proposed methods is evaluated using standard handover metrics, encompassing the number of handover attempts, successful handovers, the probability of link failure and HO delay, the occurrence of ping-pong HO, and the prevention of unnecessary handovers. Figure 4 shows the evaluation of the proposed and traditional techniques for the probability of handover issues.

The test was conducted considering 30 SCs and 20 UEs moving at different speeds. Traditional handover protocols in 5G V2X networks often depend on preset thresholds for characteristics such as SINR or RSSI, which result in less-than-ideal performance under different circumstances. They are unable to adjust their priorities in accordance with the demands of individual applications, which can lead to subpar services for such applications. The CAP mechanism, on the other hand, offers a more sophisticated and adaptable method. It guarantees that the handover decision is ideal for the present context by dynamically modifying the weights of the connection quality factors and using a multi-criteria decision-making process. This improves the overall network performance and user satisfaction. This method works especially well in V2X circumstances where application needs and network conditions can change quickly.

It is evident that CAP outperforms the traditional method, resulting in minimal HO issues. Consequently, it is a better option than conventional handover systems and produces increased service quality, increased efficiency, and improved overall user experience.

The performance of the proposed method is also compared with that reported in [8], as shown in Figure 5. The ACAVHD method serves as a context-aware state-of-the-art method for HetNets that is utilized to evaluate and compare the performance of the CAP mechanism. The basic traffic class for handover calls (Audio, Video, and General classifications) is considered in the ACAVHD algorithm. The AR, data offloading and Relay traffic profiles were added to ensure the fairness of the method comparison. It displays the average of each user-connection traffic class obtained during the simulation of 2000 different running iterations while maintaining the number of UEs and SCs per sector at 20 and 30, respectively. The UE velocity varies from low to high. The simulation results convincingly show that the CAP method significantly reduces handover issues for all the services in comparison to the ACAVHD method.

In CAP, HO attempts to show the frequency of handovers segregated for each traffic class while the UE moves from one location to another. The highest average for the number of handovers was attempted for all traffic when the UE moved at a speed of 100 km/h. Compared to other traffic classes, Relay is the most active. This is crucial for V2X to facilitate seamless connectivity. The proposed method’s impact is evident in the decreased perception of link failure and the delay of an average of 1% for all traffic types. The accepted value for ping-pong on average is 6% due to the high percentage of 84% that prevents unnecessary handover. In contrast, the ACAVHD method demonstrated low efficiency: the traffic is almost equally distributed among the services. It clearly demonstrates that the CAP method reduces the frequency of handover by nearly 99% compared with the compared algorithm. On the other hand, the CAP achieved a lower ping-pong rate, which is proportionally related to the prevention of unnecessary handovers, where the higher the number, the better. This is a result of avoiding unwanted handovers (necessary HO) as an implication of the high-ping-pong HO. This phenomenon can cause high signaling in a network and inefficient resource utilization, which, in turn, affects network efficiency. The inclusion of application-specific parameter-weighted factors in the TOPSIS method guarantees that all the user information context contributes to the HO decision, thereby enhancing the operation of services’ traffic.

Figure 6 shows that the CAP approach demonstrates efficient and consistent performance despite variations in the UE velocity, leading to the enhancement of user expectation compared to the alternative method, ACAVHD, as demonstrated in Figure 7. The hybrid channel allocation strategy in ACAVHD prioritizes handover calls and can enhance the likelihood of successful HOs. On the other hand, the static and dynamic contextual information can reduce the number of unnecessary handovers, enabling the UE to only initiate handovers when needed. However, this combination can negatively impact network performance, causing significant delays in the HO decision-making process, particularly in UDN. This delay can force the detached UE to return to the previous cell, thereby increasing the phenomenon of pin-pong HOs.

Further evaluation was conducted to measure the overall performance of the two methods, as shown in Figure 8. In these simulations, the number of UEs was fixed at 10, moving at a velocity of 20~60 km/h, while the SC density varied from 30, 50, and 100. The number of attempts at handover in ACAVHD was slightly higher than that in CAP, leading to a higher average of success (HO). It is markedly more efficient than traditional approaches. ACAVHD optimizes the selection of SCs based on a combination of mobile node speed and location. On the other hand, the CAP method effectively reduces frequent HO by employing two techniques: restricting cell scanning to only the user’s movement direction and filtering the ranking list to include only the highest PI. This method yields greater benefits when the network density is very high. Although both HO strategies ensure that the UE will not encounter interrupted services by ensuring sufficient resource allocation ruling and the enhancement of its capability to minimize link failures, the CAP shows better performance than the other method as a result of checking the cell load prior to processing a handover. The practice in the ACAVHD method of switching the connection to wait in queue for channel allocation can potentially impose an additional delay in comparison to the CAP, as the waiting time for the handover is prolonged. In addition, the UE is prohibited from proceeding with cell scanning until the consistency of the information elements stored on the central server is checked.

Ping-pong HO in ACAVHD is comparatively higher than the CAP method by approximately 30%. In ACAVHD, a handover call should satisfy two conditions before handover execution, which adds more complexity to the handover decision-making process, especially for high-mobility users. For instance, if the handover call demands a channel and the channel queue is already at the maximum capacity, then it is necessary to perform a handover to another neighbor network.

There is a chance that the channel queue is created before HO completion and that the current channel provides better services, which results in ping-pong and, therefore, unnecessary handover. In the proposed method, the inclusion of ST in the proactive handover decision prevents the UE from triggering handover when the current cell can provide a sufficient service, thereby reducing needless handover by 84%.

Additional discussions to strengthen the CAP’s performance over the ACAVHD method are detailed in the following points:

• CAP ranks SCs through TOPSIS using network performance parameters. This allows for a more precise and context-aware selection of SCs that are best suited for the application/service requirements of UEs. This minimizes resource contention and selects only the best SCs with low interference, better signal quality, and less load, resulting in high connection quality and better HO success.
• CAP includes the stay time, which determines how long a UE remains in an SC before attempting an HO. This prevents fast-moving UEs from transferring to HO frequently, thereby preventing unnecessary HOs and a ping-pong effect. CAP clearly maintains HO stability across a wide range of UE velocities.
• CAP evaluates Current SC loads before allowing HO. This is a form of proactive load management that load-balances a network and prevents HO failures due to congestion.
• Using stay time reduces transient or unstable HOs.
• CAP implements a policy-based approach to adjust weights according to the application requirements of a UE (context). This optimizes HOs to not rely only on signal metrics, but also on QoS criteria to ensure better user experience and session continuity. Using static weights will not allow the adoption of dynamic network conditions.

5. Conclusions

To conclude, this study presents a novel handover approach for 5G Vehicle-to-Everything networks by utilizing a multi-criteria decision-making methodology and considering both the application-specific quality of service needs and conventional performance measurements. The implications of the integration of AR in the context of V2X are discussed, and the proposed technique greatly improves the effectiveness and dependability of network handovers. Providing seamless connectivity, this approach minimizes network signaling overhead and maximizes resource efficiency by reducing the frequency of needless handovers by incorporating the concept of stay time. The simulation results suggest that the proposed handover mechanism performs better than the traditional method and ACAVHD, particularly in settings with dense network deployment and high mobility. The technique not only lowers unwanted handovers but also guarantees a better quality of connectivity that is specifically designed to meet the rigorous requirements of contemporary V2X applications. The findings provide noteworthy results, which indicate that this handover mechanism may be modified for use in a wider range of high-mobility scenarios outside V2X communications.

Future work can concentrate on scalability and the use of cutting-edge technologies, such as deep learning and Recurrent Neural Networks (RNNs), for performing predictions based on short-term trends. This can be used to predict UE trajectories, ST, and signal degradation, among other factors, before any HO decision is made. RL can also be integrated to optimize HO policies based on rewards. RL can adapt to dynamic network conditions and UE requirements to enable CAP’s self-optimization. In terms of further 5G development, such as 5G and Beyond and 6G, CAP can be extended to respond to the requirements of enhanced multimedia, such as holographic communications and the use of Terahertz (THz) bands.