Proficient routing by adroit algorithm in 5G-Cloud-VMesh network

Vehicular Ad hoc Networks (VANET) are pervasive, ubiquitous, and infrastructure-free without any centralized authority. VANET is a decentralized network where each node is proficient to forward data packets between the nodes dynamically [1‐4]. In VANET, a node can be bus, tram, subway, train, or taxi. Vehicles get a broadband wireless access by using heterogeneous wireless technologies, namely IEEE 802.11, IEEE 802.16 (WMAN, WiMax), 3GPP Universal Mobile Telecommunications System (UMTS), and Long-Term Evolution. A wireless mesh network (WMN) is a multi-hop communication network in which the nodes are self-organized (i.e., without a central coordinator) [5]. VMesh is a hybrid network, which combines the features of VANET and WMN. VMesh network is the communication environment between the vehicles made up of radio nodes organized in a mesh topology. VMesh generates a loop among the moving vehicle. In the loop, we can add, update, or remove the nodes [6]. Vehicles are becoming more sophisticated with powerful on-board computing capabilities such as tons of on-board storage and short- or medium-range wireless transceiver with no power limitations. Advancement in wireless networks crafts the access to the Internet from vehicles. The high-flying industries and governments such as Audi, BMW, Daimler-Chrysler, and Toyota have commenced important research for vehicle to vehicle communications. California Partners for Advanced Transit and Highways, Crash Avoidance Metrics Partnership, FleetNet, Advanced Driver Assistance Systems, Chauffeur in EU, CarTALK 2000, and DEMO 2000 are petty distinguished research, which are moving towards the realization of intellectual transport services.

With surging vehicle Internet services, VMesh network is facing unique opportunities and challenges. To coup the rapid development and demands due to the multimedia services, social networks, and entertainments, it is far from enough to only expand and upgrade current networks, which results in a substantial increase in CAPital EXpenditure (CAPEX) and OPerating EXpenditure (OPEX) without reasonable investment returns. Therefore, in order to offer a High-Quality of Experience (HQoE) for vehicle users with affordable costs, new network architecture is designed to enable sustainable network evolution to the 5G. Till now, much attention has been paid to emerging 5G technologies ranging from information processing to networking techniques. Most of the focus has been on further improving radio access efficiency through advanced signal processing, e.g., cooperative relaying, massive MIMO, and millimeter-wave communications [7‐13]. On the other hand, the existing shaft-like network structure with centralized data processing is not suitable for flexible and diverse VANET Internet services, which undoubtedly restricts the service provisioning ability of the VMesh network. Hence, designing new network architecture becomes imperative. This motivates us to propose a new architecture for VMesh network with the integration of cloud and 5G technologies. However, there is still much work that needs to be done to improve the 5CVN architecture with good backward compatibility. This paper emphasizes how to maximize the utilization of resource without delay and congestion and to smoothly evolve towards converged 5CVN.

The motivations of the paper are as follows, nowadays, users of all ages want to share and get the information at anytime and anywhere via vehicle Internet service. Also, the growth of the Internet-of-Things (IoTs) has had serious effects in the VANET performance. It is expected that by the year 2020, average user data rates will increase from around 20 MB/s to 1 GB/s, and connection density will also increase from 140,000 connections per square kilometers to 6 million connections per square kilometers. Moreover, new applications such as the Internet-of-Vehicles are envisioned for 5G, which will demand much shorter latency from the underlying network. Considering the existing network infrastructure, telecommunication operators need to invest heavily on both network maintenance and upgrade to support ever increasing capacity requirements. However, revenues from increased data traffic cannot keep up with the investment. How to reduce CAPEX/OPEX and create new revenue flows are crucial issues facing telecommunications operators. The proposed 5CVN-adroit algorithm may perhaps solve the above issues.

In this paper, an integration of 5G-cloud-VMesh network is introduced. The envisioned architecture shall enable data access from the vehicles, anytime and anywhere. In particular, the integration of IEEE 802.11-based multi-hop VANETs with 5G shall contribute to the evolution of beyond 4G wireless communication systems. The proposed 5CVN-adroit algorithm addresses the choosing of an optimal gateway and Aggregator Node B (ANB) in the network layer from and to the service requesters and the cloud using 5G technologies. 5CVN-adroit algorithm consists of two key processes such as metric-based gateway selection and metric-based ANB selection. In the majority of the existing research, gateways are considered static and deployed on the road at fixed distances from each other, depending on their wireless transmission range, which makes the deployment costlier. Moreover, the multi-hop and dynamic nature of VANET communication impacts the stability of the links to these gateways. Furthermore, as these gateways are fixed, the discovery and routing processes are mainly pro-active. Although the pro-active routing mechanisms reduce delay, they increase the frequent changes in the predefined routing tables of vehicles and routing overhead. To deal with these limitations, this paper defines the concept of vehicle gateways. Any vehicle in the VMesh network can act as a gateway. A gateway refers to the dual-interfaced vehicle that relays the packet from sources to the 5G backhaul network. It is enabled with dual interfaces of IEEE 802.11p and the 5G backhaul networks. The major challenge is to amalgamate these two network interfaces on a gateway, as they lie in two different spectrum regions. The next challenge is to choose a minimum number of optimal gateways using relevant metrics. There could be several nodes in the VMesh network, which possess the important criterion and required metric information, and hence, the qualification to serve as gateways. The focus of this paper is on defining a mechanism that chooses a minimum number of optimal gateways, at an instance, so as to avoid the bottleneck at the 5G backhaul network. With this regard, as pro-active gateway discovery reduces delay and reactive discovery reduces routing overhead, this paper envisages a hybrid gateway discovery mechanism for VMesh network, combining the pros of both pro-active and reactive concepts.

For example, network statistics show that 70 % of data traffic is carried by 30 % of base stations. To solve the above issue, we introduced ANB pool; a large number of ANBs can form an “ANB pool” with aggregated processing capabilities to serve the traffic requirements of the coverage area. For instance, there are certain regions with heavy traffic loads and certain ANB are overloaded. In such cases, ANB selection has been introduced to select the optimal ANB among the ANB pool. In the majority of the existing research, conventional centralized cloud is considered. However, the conventional centralized cloud may result in a high delay of offloading data delivery and in congestion of backhaul due to the transmission of the offloaded data. The perceived delay caused by the centralized cloud can be minimized by adopting a distributed cloud deployed over ANB of cellular network.

Our paper deals with the above-identified issues by the following key contributions such as:

The rest of this article is organized as follows. In Section 2, we discuss the proposed 5CVN architecture and its main features. In Section 3, we describe the 5CVN-adroit algorithm. In Section 4, we discuss and analysis the performance of the 5CVN-adroit algorithm. Finally, in Section 5, we present our conclusions and a description of future work that needs to be done.

To facilitate the flexible network control and reduce the network capital and operational expenditures, the complimentary concepts, 5G software-defined networking (SDN) components, and cloud have been integrated with VMesh networks. Through the proper cooperation between newly developed 5CVN and radio access network, vehicle user can enjoy a much better quality of experience with lower access delay and congestion. 5G SDN is fast evolving and becoming the focal point and a growth area in networking. 5G SDN is a new paradigm in networking, e.g., data and control plane separation, globally optimized network management, and centralized network controller [14‐16].

Figure 1 illustrates the 5CVN network architecture. The 5CVN is heterogeneous and IP-based model with relays. The architecture includes a number of independent vehicles with the autonomous radio access technologies, ANB, MME, Serving Gateway (SG)/Gateway GPRS Support Node (GGSN), SDN controller, and cloud server. VMesh network user’s vehicle may have any technology provisions. ANB has local cloud and it also has a computing capability for offloading a computation from the vehicle. The local cloud consists of the service which the user opts often. ANB provides the air interface between the control plane protocol terminations and the user plane towards the autonomous Human Machine Interface (HMI). The interfaces of network are built on IP protocols. ANB is equipped with different MAC protocols to deliver different types of data information. The SG/GGSN act as a limited anchor for the mobility service for transmitting and receiving packet rates from and to the requester and cloud to serve the HMI request. Mobility Management Entity (MME) is an entity to afford signaling only, and later, the user packets of the IP do not pass over the MME. SDN controller allows administrators to manage network services through abstraction of lower-level functionality. A cloud server is a logical server that is built, hosted, and delivered through a cloud computing platform over the Internet. Cloud servers exhibit and possess similar capabilities and functionality to a usual server but are accessed remotely on-demand from a cloud service provider. The cloud server provides computation, data access, software, services, and storage that do not require service requester’s knowledge about the physical location and configuration of the system that deliver services.

As shown in Fig. 2. Service-level agreement (SLA) is the negotiation between the cloud server and service requesters for maintaining the QoS. Some metrics that SLAs may specify in the 5CVN are as follows:

The 5CVN-adroit algorithm discovers an optimal gateway and ANB in the network layer to route the request/response from/to the requesters and the cloud using 5G technologies. An ingress vehicle (service requester vehicle) coverage area is IEEE 802.11p wireless transmission range. As expressed in Eq. 1, the coverage area of an ingress vehicle is determined as follows:

where TR_max denotes the maximum IEEE 802.11p transmission range and α reflects the wireless channel fading conditions of the ingress vehicle location. For instance, α can be set to small values in environments when there are no major obstacles and takes high values in urban areas with tall buildings. A mapping function between the geographical locations and the values of α can be provided by using position system such as GPS and Galileo. As illustrated in Algorithm 1, the 5CVN-adroit algorithm consists of two key processes such as metric-based gateway selection and metric-based ANB selection.

We implemented the 5CVN-adroit algorithm and the existing techniques on network simulator (NS2) and used the topology of the network generated by Simulation of Urban Mobility (SUMO). SUMO is the microscopic, open source, and multimodal traffic simulator. The simulation allows to address a large set of traffic management topics. Each vehicle is modeled explicitly and has an own route and travel independently through the network. Here, SUMO is coupled to a communication network simulator using a middleware (VSimRTI). The V2X Simulation Runtime Infrastructure (VSimRTI) is a framework which enables the groundwork and execution of V2X simulations. The easier integration and exchange of simulators enables the replacement of the most relevant simulators for a realistic presentation of traffic, communication, and the execution of V2X applications. The NuoDB is a cloud database and offers a well-off SQL implementation and factual ACID transactions. Table 1 illustrates the characteristics of the environment in which the simulation is experimented. In the simulation, the request generation speeds remnants at five packets per second when the vehicle density changes.

In our existing work, we have proposed Optimized Heuristic Buffer-based Routing (OHBR) which focuses on proficient routing based on the buffer size and heuristic fitness function, in order to reduce the packet loss, retransmission, gridlock, protocol overhead, and delay in VMesh milieu [5]. In this work, we do not consider the base station selection which results in data traffic at the base station. The data traffic results in delay, packet loss, and packet collisions. Abderrahim et al. proposed a mechanism and named as Clustering-based Multi-metric adaptive mobile Gateway Management (CMGM) in which vehicles are dynamically clustered according to different related metrics such as UMTS received signal strength, direction of vehicles, and vehicle to vehicle distance [17]. However, the cluster mechanism leads to confusion in electing the cluster head. And as well in CMGM fixing, the gateway candidate progression is complex which results in a delay in gateway discovery mechanism. Kaouther et al. have proposed efficient and scalable LocVSDPs for VNs. The proposed protocols have vehicles provided with an efficient mechanism to locate service providers and how to reach them (routing) simultaneously [18]. However, in LocVSDPs for VNs, clustering leads to confusion in electing the cluster head and it will increase the routing overhead and delay. The performance of the proposed 5CVN-adroit algorithm and existing OHBR, 3G-CMGM, and LocVSDPs are assessed in terms of the packet delivery ratio, average response time, routing overhead ratio, and packet collision ratio. The intricate descriptions of these indexes are as follows:

Table 2 depicts the simulation results obtained by the proposed 5CVN-adroit algorithm (A), existing OHBR (O), 3G-CMGM (C), and LocVSDPs (L) for different request generation speed (RGS) and vehicle density (VD). Based on the obtained results, we calculate the PDR, ART, RO, and PCR. In our existing work Optimized Heuristic Buffer-based Routing (OHBR), we do not consider the base station pool and the base station selection which results in data traffic at the base station when the RGS and VD rises. This results in delay, packet loss, and packet collisions. Consequently, PDR and RO decrease, and ART and PCR increase. In existing 3G-CMGM, the cluster mechanism leads to confusion in electing the cluster head. It results in a decrease in the PDR and RO and increase in the ART and PCR. In addition, in CMGM fixing, the gateway candidate progression is complex which results in a delay in gateway discovery mechanism. The existing LocVSDPs relies on roadside routers. The roadside router has to be planned and deployed on the road side. Consequently, the deployment cost is high. LocVSDPs is also the cluster-based approach. The cluster creates confusion in the routing process, results in the diminishing of the PDR and RO, and rises in the ART and PCR. However, in the proposed mechanism by using the 5G technology components, adroit algorithm, and local cloud at ANB, the VMesh user can enjoy with the minimum delay and maximum delivery ratio.

The impact of various RGS on PDR is illustrated in Fig. 3. It can be deduced from this figure that PDR decreases as the RGS rises. Higher RGS causes more network topology changes and more link failures and thereby PDR decreases accordingly. Owing to the introduction of gateway and ANB selection based on the relevant metrics, the probability of disconnections between platoons shows an improvement in the 5CVN-adroit algorithm over to the OHBR, 3G-CMGM, and LocVSDPs. Hence, the PDR of the 5CVN-adroit algorithm outperforms the existing techniques. As depicted in Fig. 4, the ART of OHBR, 3G-CMGM, LocVSDPs, and 5CVN-adroit algorithm increases when the number of request increases. Higher RGS generates more network topology changes and more delay. In existing techniques, the selected gateways and the base station will be overloaded with packets, some of which will be left over, finally impacting the packet loss and retransmission, which will increase the response time. In the 5CVN-adroit algorithm, as the gateway and the ANB are chosen based on the minimum number of packets in the buffer, it will decrease the processing delay. In addition, in the 5CVN-adroit algorithm, the next-hop gateway and the ANB are picked based on the fitness value and reserved ahead before ingress vehicle transmitting the request. As a consequence, the ART of the 5CVN-adroit algorithm outperforms the ART of OHBR, 3G-CMGM and LocVSDPs. Accordingly, we have compared the ROR performance of the 5CVN-adroit algorithm, OHBR, 3G-CMGM, and LocVSDPs with RGS varying in Fig. 5. It can be figured out from this figure that ROR decreases as the RGS rises. By virtue of gateway selection, the 5CVN-adroit algorithm is more likely to find a route from source to destination without any confusion in the routing process. Thus, the 5CVN-adroit algorithm outperforms the ROR of OHBR, 3G-CMGM, and LocVSDPs in terms of higher routing overhead ratio. It can be deduced from Fig. 6 that PCR of OHBR, 3G-CMGM, LocVSDPs, and 5CVN-adroit algorithm increases as the RGS increases. This result is rational and illustrates that high data rate may easily inundate the service channel of IEEE 802.11p since no differentiated service is being used. Compared to OHBR, 3G-CMGM, and LocVSDPs, the 5CVN-adroit algorithm shows the superlative PCR owing to its routing selection concern which combines the delay estimation and path connectivity together.

The influence of vehicle density on PDR of the 5CVN-adroit algorithm, OHBR, 3G-CMGM, and LocVSDPs is plotted in Fig. 7. It is worth stressing that the impact of vehicle density on PDR is more apparent than that of RGS. The PDR of OHBR, 3G-CMGM, LocVSDPs, and 5CVN-adroit algorithm generally shows a diminishing trend in the growth of vehicle density. However, a large vehicle density will introduce more collisions and retransmissions on the IEEE 802.11p MAC layer. Thus, the OHBR, 3G-CMGM, LocVSDPs, and 5CVN-adroit algorithm experience a dropping PDR when vehicle density increases. Due to the usage of real connected paths in 5CVN-adroit algorithm between source and destination, it can easily tolerate MAC collisions, thus still providing better PDR when vehicle density is more. From Fig. 8, it can be deduced that with the continuous increase of vehicle density, the ART of OHBR, 3G-CMGM, LocVSDPs, and 5CVN-adroit algorithm also increases. Higher density generates more network topology changes and more delay. By selecting the qualified gateway and ANB based on relevant metrics, ART of adroit algorithm outperforms the ART of existing techniques. In addition, by using the local cloud in computing enhanced ANB, the ART of 5CVN-adroit algorithm outperforms the existing techniques. As shown in Fig. 9, the ROR of the 5CVN-adroit algorithm, OHBR, 3G-CMGM, and LocVSDPs decreases when the vehicle density increases. The existing techniques create confusions in the routing process; hence, it will decrease the routing overhead ratio. To avoid the confusion in adroit algorithm, quantified gateway and ANB has been selected. In addition, the next-hop gateway and ANB is picked and reserved ahead; that causes less routing overhead. Therefore, the ROR of the proposed scheme is high compared to the existing techniques. The 5CVN-adroit algorithm has some advantage in the aspects of ROR by considering important path selection factors such as link availability and link quality, which is of great importance to select a better route. Thus, the 5CVN-adroit algorithm shows better reliability and scalability. PCR is affected by different parameters such as vehicle density and traffic load. The results presented in Fig. 10 suggest that the proposed 5CVN-adroit algorithm is not affected as much as existing techniques. The increase in vehicle density brings more collisions on MAC. The reason behind is that more retransmission and collision causes the routing layer to use more beacons. The 5CVN-adroit algorithm reduces the retransmission and collision by introducing an ANB selection mechanism among an ANB pool. Thus, the PCR of adroit algorithm outperforms the existing techniques.

The proposed 5CVN-adroit algorithm affords high performance, globally optimized load balancing and network usage increased from 30 % to over 90 %. In this paper, we initiate an architecture that amalgamates 5G SDN, cloud, and VMesh networks, which offers advanced network services and apps at software speed for the users. The simulation results show that our 5CVN-adroit algorithm is very suitable for real-time applications and outperforms the existing techniques on packet delivery ratio, average response time, routing overhead ratio, and packet collision ratio. However, there is still much research needed to fully exploit the potential of 5CVN. This paper may be giving a good platform to motivate the researchers for a better outcome of different types of problems in next generation networks.