Beyond 5G and future next-generation networks will have to cope with the ever-growing traffic demand for mobile traffic, as well as low-latency communications. Network densification has been long proposed as a solution for augmenting the available wireless links with more technologies, thus enhancing the available capacity for the end-users. Nevertheless, selecting the optimal split of traffic among the available links is not a trivial decision. Machine Learning (ML) approaches can assist in these decisions, by forecasting metrics collected directly from the RAN, towards predicting the near-future performance, and appropriately selecting the split of traffic. In this work, we evaluate a total of 22 different ML models in such a traffic steering use case, towards determining the solution that yields the best results in terms of accuracy of predictions, training time, and computational resources. We use a real-world testbed prototype based on OpenAirInterface to evaluate our contributions, and use realistic mobility datasets for emulating client mobility. Our results show that the different algorithms can present variations in terms of the achievable throughput, but several can substantially improve the offered wireless network capacity.

As networks become denser and more heterogeneous different paths can be considered in order to reach each multi-homed UE, offering optimal performance. 5G and beyond networks feature contributions related to the dynamic programming of the network, from the operator side, in order to optimally allocate resources in the network. In this work, we consider such a case, where network access is provided to the end-users via heterogeneous (3GPP and non-3GPP) Distributed Units (DUs), converging to a single Central Unit (CU), and programmable on the fly with external interfaces. We employ Machine Learning (ML) methods in order to forecast the Quality of Service (QoS) that a wireless client will get from the network in the near future based on the Channel State Information (CSI) metric. Subsequently, we appropriately steer the traffic over the different heterogeneous DUs for ensuring that the network meets the needs of the UEs. We design, develop, deploy and evaluate our method in a real testbed environment, using emulated mobility. Our results show that the overall throughput of each UE can be drastically improved compared to existing allocation mechanisms.

5G redefines the network architecture, through the introduction of base station disaggregation, adding up to the overall flexibility for deployments of network elements based on the existing demand. Moreover, through the wide application of Multi-access Edge Computing (MEC), the latency for accessing services offered on top of the infrastructure can be drastically minimized, thus supporting several applications that exchange critical data over the network. Vehicular communications are a type of applications that can highly benefit from the MEC approach, as they need to exchange critical data with roadside infrastructure (V2I). Nevertheless, the technology through which the vehicles exchange data with the infrastructure may cause fluctuations to the network performance observed by the end-users. In this work, we blend the concept of disaggregated 5G base stations with integrated access for non-3GPP technologies, and the MEC concept, by realizing a novel placement for services that is not yet suggested in existing literature. The setup is applied to a vehicular environment, providing the ability to dynamically steer the traffic over multiple technologies serving each user (vehicle) of the network. We implement, deploy and evaluate our solution in a testbed environment, providing proof on reduced latency for accessing the MEC services as well as the role that the technology selection plays for the overall performance observed by the end-users.

Mobile telecommunication infrastructure has been considered as the deployment of predefined network components at the end-user access. The advent of novel 5G technologies allows the re-conception of the traditional model through which network services are offered to network User Equipment (UE) devices, through the introduction of new network architectures. Cloud-based Radio Access Networks (Cloud-RANs) add to the network flexibility, allowing on-demand instantiation of new base stations. The introduction of heterogeneous technologies in the cells, and the fact that mobile devices are equipped with diverse wireless interfaces create fertile ground for the operators to select the technologies through which each UE in a cell is served, towards maximizing the perceived QoE. In this direction, we apply the Portfolio Theory, stemming from the finance field, to reveal that the appropriate combinations of networking technologies and traffic allocation can achieve the demanded throughput. In this work we provide a detailed description of this architecture and apply the Portfolio Theory for selecting the appropriate combinations of radio access technologies, towards achieving minimum volatility in throughput performance of a heterogeneous Cloud-RAN. We investigate our approach with proof-of-concept testbed experiments that showcase the applicability and performance of Portfolio Theory in real wireless environment conditions.

Multi-access Edge Computing (MEC) is expected to play an important role in next generation networks, as it is able to provide resources accessible through multiple wireless technologies located close to the network edge. MEC is therefore an enabler for low-latency applications, allowing novel time critical services to be offered through mobile networks. Nevertheless, hosting multiple service providers over the same physical infrastructure shall carefully consider the needs of the MEC enabled applications. Moreover, different suggested placements for the MEC service provide fertile ground for the differentiation of the hosted service providers. In the meantime, the integration of multiple technologies in the wireless access of the MEC concept is increasing the complexity for efficient allocation of network and computational resources among the involved stakeholders. In this work, we model the MEC resource allocation problem for different service providers by using a pricing scheme. We consider two different deployments for the MEC services when considering a multi-technology Cloud-RAN base station: either at the fronthaul interface or located at the Core Network. We integrate intelligence to the MEC enabled framework, by considering the available links through which each client is served. We employ testbed experimentation in order to illustrate the efficiency of our scheme and demonstrate how we achieve efficient allocation of the MEC resources and wireless technologies used for the system under consideration.

In this paper, we focus on 5G heterogeneous networks, considering the existence of multiple Distributed Units (DUs) that can provide access to end users implementing several access technologies, managed by a Central Unit (CU) responsible for the allocation of network resources. Based on a distributed dynamic pricing scheme, that gives to User Equipment (UE) the ability to select the appropriate Radio Access Technology (RAT) depending on its sensitivity to congestion, we investigate a scheme of greater granularity, where UEs are able to allocate each of their traffic classes to the appropriate RAT, exploiting their multi-homing features. As UEs are sequentially polled to request for network resources, we develop a centrally controlled proportionally fair ranking as a benchmark policy. We then propose a dynamic polling policy that presents close performance to the benchmark policy, while maintaining a distributed nature. We evaluate our framework for a variety of traffic classes in terms of Quality of Service (QoS) requirements, and we provide results on capacity utilization and load distribution over available RATs, as well as access price variations.

Multiple-access Edge Computing (MEC) has been proposed as a means to minimize the user to service path latency, by deploying and operating datacenter resources close to the network edge. The introduction of 5G mobile network services, and their provisioning through disaggregated base stations complying with the Cloud-RAN paradigm, allows the redefinition of the traditional Edge Computing by offering deployment of services even closer to the network access edge. In this work, we leverage a disaggregated heterogeneous 5G infrastructure, compliant with the 5G New Radio (NR) specifications, and present a scheme for placing the services even closer to the Edge, close to the concept of fog computing. We develop a scheme for the OpenAirInterface platform that allows services to be executed close or over the machines hosting the radio services for the network access. By exploiting features for integrating heterogeneous radio resources in the cell, we are able to create a controller interface for selecting the optimal radio access technology used to serve each user of the network from the MEC service perspective. We evaluate our solution in a real testbed setup, and measure performance related indicators for our solution by using adaptive video streaming. Our results illustrate up to 80% better video qualities delivered to the end user when appropriately selecting the access technology.

Radio Access Network (RAN) disaggregation is transforming mobile networks, as it allows the single click instantiation of base stations in the Cloud, and can potentially ease the integration of heterogeneous technologies for wireless access. At the same time, Multi-access Edge Computing is proven to minimize the latency for accessing services located at the network edge. In this demo, we showcase a disaggregated heterogeneous Cloud-RAN, consisting of cellular and WiFi infrastructure, providing access to edge resources placed on the fronthaul of the network to multi-homed UEs. We experiment with different wireless technologies and placements for the MEC service, and illustrate our experimental results.

Ultra-Dense Heterogeneous Network deployments are expected to boost the offered network capacity and enhance the user-perceived Quality of Experience, through the simultaneous offering of multiple technologies using distinct or shared wireless spectrum. In such environments with a plethora of available Radio Access Technologies (RATs), the network UEs shall decide either independently or assisted through operator based services on which network they shall use to better serve their needs. In this work, we model the network selection problem in a Multi- RAT system, based on the Paris Metro Pricing (PMP) scheme, enhanced with dynamic pricing formed by the congestion of each available technology. We assume that the network UEs are equipped with multi-homing features and thus are able to use concurrently more than one technologies, based on the requirements of the applications requesting network connectivity (e.g. UHD video streaming). We port and experimentally evaluate the proposed system model over a testbed setup, using distributed components running at the UEs and at a Core Network controller. We provide evaluation results on the average cost per UE for the selected RATs, the average data rate distribution of each UE per RAT and how these performance metrics are affected under different client ordering policies at the network controller.

Heterogeneous network access has been proposed as a solution for the continuously deteriorating congestion problem of cellular infrastructures. In this paper we focus on a multi- Radio Access Technology (RAT) environment and we provide a solution based on the Paris Metro Pricing (PMP) scheme, which was first applied in Paris metro. The concept of this pricing scheme was to provide differentiated service classes for customers who desired to avoid being congested, while maintaining the same wagons, characterized only by a different ticket price. The proposed solution extends the classic PMP policy by inducing dynamic prices formed by the congestion of each available technology. We investigate the performance of our dynamic PMP scheme taking into consideration a mobility model of the interested users. Through simulations and testbed experimentation we provide evaluation results on the average throughput, the acceptance capability of the incoming users and how these performance metrics are affected under different mobility conditions.

Cognitive radio systems have gathered a lot of research interest during the last decade. Accuracy of spectrum sensing and efficiency of free spectrum utilization are considered as the primary objectives in this emerging technology, which promises a boost in wireless network performance, through exploitation of underutilized licensed frequency bands. As the focus of researchers is usually on these two major challenges, other aspects have been in part underestimated. In this work, we consider two factors that are rather important for evaluation of cognitive platforms, namely sensing delay and energy efficiency. The first is related to the latency induced by the spectrum sensing process and its impact on sensing efficiency, which is tightly connected to both the QoS performance of secondary users and the protection of primary users. On the other hand, energy consumption is considered as a crucial issue in all types of wireless communications, due to restricted battery autonomy of mobile devices, as well as for moving towards "greener" solutions in telecommunications. Therefore, it is important to extend existing testbed experimentation tools and develop new ones, in order to equip cognitive testbeds with such advanced monitoring capabilities. In this work, we present a monitoring procedure that has been directly integrated in the experimentation tools of the CREW testbed federation and demonstrate how it aids in the online evaluation of four different cognitive platforms in terms of the aforementioned metrics.