Papers by Marius C Vochin
Towards Open RAN in beyond 5G networks: Evolution, Architectures, deployments, spectrum, prototypes, and performance assessment
Computer Networks, 2025
Along the evolution of mobile communication systems during the past decades from 1st generation (... more Along the evolution of mobile communication systems during the past decades from 1st generation (1G) to 5th generation (5G), radio access networks (RANs) have been traditionally implemented and deployed based on vendor-specific hardware and proprietary software. In recent years, there is a surge of interests and efforts among mobile network operators (MNOs) in developing RAN solutions based on open interfaces to facilitate flexible operations and versatile functionalities of RANs using reconfigurable hardware and inter-operable software. In the meantime, various industrial and academic stakeholders are keen on developing open-source driven RAN solutions based on commercial-off-the-shelf (COTS) hardware like software defined radio (SDR). In this article, we convey a comprehensive and systematic study on Open RAN technologies spanning from the theories (from RAN evolution to Open RAN architecture) and the concepts (from disaggregation and functional split to Open RAN architectures), to commercial and prototype deployments as well as implementation platforms. In addition to an in-depth comparison between the 3rd generation partnership project (3GPP) and Open RAN Alliance (O-RAN) interfaces and functions, we shed light on the spectrum allocation aspects for Open RAN deployment covering both public and private 5G networks. Furthermore, this article presents two prototype 5G networks that we have developed based on COTS hardware and open-source software suites as well as the configuration and performance of these prototype proof-of-concept on-campus networks.
Enabling Reinforcement Learning for Network Slice Management in Multi-Agent 5G Networks
To enhance dynamic resource adaptation in fifth generation (5G) networks, network slicing managem... more To enhance dynamic resource adaptation in fifth generation (5G) networks, network slicing management empowered by artificial intelligence (AI) through decision-making algorithms may improve resource utilization, quality of service (QoS), as well as network scalability and flexibility. In this paper, we propose an AI-driven network slice management (AI-NSM) framework that enables enhanced adaptive resource allocation for 5G networks by ensuring additional management and orchestration for network slices. The integration of AI-NSM into 5G networks exhibits superior adaptability supporting dynamic organization of network slices based on predicted traffic patterns through reinforcement learning (RL), leading to reduced latency, optimized resource allocation, and improved QoS. Based on a virtualization platform through Oracle virtual machines, we implement an AI model including a multi-agent deep deterministic policy gradient RL algorithm that provides complementary support for other network slice management functions. Through implementation and experiments, we demonstrate that AI-NSM can enhance resource allocation and improve network responsiveness for slicing in 5G networks.
Campus-Based Full-Scale and Portable Open-Source 5G SA Networks: Prototyping and Experiments
In light of the thrilling development of future mobile communication systems and worldwide 5G dep... more In light of the thrilling development of future mobile communication systems and worldwide 5G deployment, there is a trend toward open-source-based solutions, which could reduce costs, provide flexibility, enable innovation, and expand knowhow on 5G RAN and 5G CN operations. In this study, we design and implement two distinct 5G SA open-source-based testbeds. The first network architecture, a Portable Network, comprises a Lenovo laptop used in tandem with a USRP B210. The second network architecture, a Full-Scale Network, is a campus-based 5G network deployed at a lab facility, consisting of a bare-metal Dell server and a USRP X300. Based on the implementations of these two network architectures, extensive experiments have been performed with a COTS UE connected to the network. Our results provide insight on the performance of such networks with respect to network connectivity and computing performance.
Applied Sciences, 2024
This paper illustrates a general framework in which a neural network application can be easily in... more This paper illustrates a general framework in which a neural network application can be easily integrated and proposes a traffic forecasting approach that uses neural networks based on graphs. Neural networks based on graphs have the advantage of capturing spatial–temporal
characteristics that cannot be captured by other types of neural networks. This is due to entries that are graphs that, by their nature, include, besides a certain topology (the spatial characteristic),
connections between nodes that model the costs (traffic load, speed, and road length) of the roads between nodes that can vary over time (the temporal characteristic). As a result, a prediction in a node
influences the prediction from adjacent nodes, and, globally, the prediction has more precision. On the other hand, an adequate neural network leads to a good prediction, but its complexity can be higher.
A recurrent neural network like LSTM is suitable for making predictions. A reduction in complexity can be achieved by choosing a relatively small number (usually determined by experiments) of hidden levels. The use of graphs as inputs to the neural network and the choice of a recurrent neural
network combined lead to good accuracy in traffic prediction with a low enough implementation effort that it can be accomplished on microcontrollers with relatively limited resources. The proposed
method minimizes the communication network (between vehicles and database servers) load and represents a reasonable trade-off between the communication network load and forecasting accuracy.
Traffic prediction leads to less-congested routes and, therefore, to a reduction in energy consumption.
The traffic is forecasted using an LSTM neural network with a regression layer. The inputs of the neural network are sequences—obtained from a graph that represents the road network—at specific moments in time that are read from traffic sensors or the outputs of the neural network
(forecasting sequences). The input sequences can be filtered to improve the forecasting accuracy. This general framework is based on the Contiki IoT operating system, which ensures support for wireless communication and the efficient implementation of processes in a resource-constrained system, and it is particularized to implement a graph neural network. Two cases are studied: one case in which the traffic sensors are periodically read and another case in which the traffic sensors are read when
their values’ changes are detected. A comparison between the cases is made, and the influence of filtering is evaluated. The obtained accuracy is very good and is very close to the accuracy obtained in an infinite precision simulation, the computation time is low enough, and the system can work in real time.
IEEE Access, 2024
Electric vehicles represent an overwhelming worldwide trend towards environment-friendly transpor... more Electric vehicles represent an overwhelming worldwide trend towards environment-friendly transportation and the number of electric vehicles has been increasing rapidly for more than one decade. However, uncoordinated charging of electric vehicles leads to increased congestion at charging stations and significantly strains the power supply grid. Recently, this problem has become more serious and it has received remarkable attention in both societies and the research community. In this paper, we present an overview of electric vehicle charging scheduling mechanisms, focusing on two main categories of schemes, namely centralized and decentralized. Specifically, the paper covers the optimization techniques that are applicable for optimizing the charging of electric vehicles. After reviewing the state-of-the-art research, we identify limitations and gaps in current solutions, such as challenges in handling large numbers of electric vehicles, over-reliance on predictive models without accounting real-time grid status, and a commonly overlooked aspect on user preferences in charging schedules. Furthermore, we shed light on potential research directions relevant to this topic. INDEX TERMS Electric vehicle, optimization techniques, charging scheduling, optimal charging scheme, charging behaviors
Good Practices and New Perspectives in Information Systems and Technologies. WorldCIST 2024. Lecture Notes in Networks and Systems, 2024
In recent years, together with advances in the software defined radio (SDR) field, the concept of... more In recent years, together with advances in the software defined radio (SDR) field, the concept of mobile private networks (MPNs) gained momentum. Several solutions based on open-source software modules and using SDR platforms as hardware were developed. In the current paper, the radio coverage of a private 4G mobile network was evaluated using the HTZ Communication software and field-test measurements. The network was implemented using an USRP B210 SDR platform and the srsRAN-4G software suite. A comparison between the coverage estimated using different propagation models and measurements is performed, together with an analysis of the obtained results.
UPB Scientific Bulletin, Series C: Electrical Engineering, Aug 12, 2021
To outcome the higher demand on IT applications, the enterprises and
service providers must buil... more To outcome the higher demand on IT applications, the enterprises and
service providers must build and expand the use of edge computing. This can lead to
possible difficulties concerning how to place workloads of network services
automation and evaluate the costs in an optimal way. This paper’s intention is to
present a scenario with three different virtual network functions designed with
specific restrictions and different compute costs and the Control System Hub with
Open Source Mano solution to limit and face those restrictions. To achieve this, the
optimal placement algorithm uses the Control System Hub network analytic
platform along with virtual charmed factors for control and Quality-of-Service
measures.
2022 International Conference on Computer Communications and Networks (ICCCN), Jul 1, 2022
Although the number of massive machine type
communications (mMTC) devices covered by a fifth gen... more Although the number of massive machine type
communications (mMTC) devices covered by a fifth generation
(5G) cell could be huge, not all devices are always active. For
radio resource allocation, it is vital that a base station has the
knowledge on the number of active mMTC devices in each
frame. In this paper, we develop a mathematical framework to
perform pseudo-Bayesian estimation on the number of active
devices within a 5G new radio (NR) subframe. The estimation
is performed by the base station on a subframe-by-subframe
basis based upon which a transmission permission probability
is imposed on active devices for their transmissions in the next
subframe. The active devices include both backlogged devices
and new arrivals in the current subframe that will attempt for
medium access in the next subframe relying on a combination of
NR radio resources in both time and frequency domains. Such
an estimation constitutes an important component for the design
of grant-free access schemes for mMTC uplink traffic.
Selecting multiple communication technologies in IoT platforms
Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies XI
The Perception layer in Internet of Things (IoT) architectures is responsible for connecting sens... more The Perception layer in Internet of Things (IoT) architectures is responsible for connecting sensor nodes and data acquisition units such that sensing devices capture relevant data from the corresponding environment. Most IoT platforms are designed to transmit data at fixed time intervals, which is a disadvantage in the modern world dominated by rapid changes in the evolution of events. This paper addresses improvements performed on an IoT platform dedicated to critical applications (e.g., fire, air pollution). This novel approach assumes the use of two equations empirically determined to compute the time interval between successive transmissions depending on the detected event. A new method for communication technology selection (LoRaWAN, Wi-Fi or cellular) is implemented and the time interval between two successive transmissions is adjusted according to the occurring event. Comparisons were highlighted for each analyzed case. The proposed method proved to be suitable for critical scenarios or scenarios that can generate false-positive alarms, due to abnormal variations of parameters.
Object Detection in Autonomous Vehicles
2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC)
In the coming years, autonomous driving will be the primary focus of the automobile industry. The... more In the coming years, autonomous driving will be the primary focus of the automobile industry. The great majority of accidents are caused by human mistakes, and autonomous cars can help to lower this number significantly, thus improving road safety. Object identification plays a critical part in autonomous vehicle driving, and deep learning techniques are used to implement it. YOLO is one of the most common methods for recognizing and identifying things that emerge on the road. Its popularity has developed as a result of its superior performance in terms of speed, high accuracy, and learning capabilities when compared to other object recognition approaches such as Retina-Net, fast R-CNN, and Single-Shot MultiBox Detection (SSD).
Building an Open Source Containerized 5G SA Network through Docker and Kubernetes
2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC)
Automating software orchestration and service development represents the newest trend in the deve... more Automating software orchestration and service development represents the newest trend in the development of fifth generation (5G) core network (CN) as it enables flexible and scalable service deployment. The building blocks for such a trend include Containers, Docker, Kubernetes, and other orchestration methods that facilitate easy scaling, management and control, load balancing, and personalized quality of service. In this paper, we develop a containerized 5G standalone (SA) network, building two types of network topologies for 5G SA deployment based on the concepts of 5G cloud network functions, Docker containers and Linux virtualization. Based on our implementation of both Minimalist Deployment and Basic Deployment, an assessment on the attach procedure is performed through next generation application protocol (NGAP) filtering along with subscriber information. Moreover, emulated transmission control protocol (TCP)/user datagram protocol (UDP) traffic is injected into the network and its performance is evaluated based on metrics such as traffic volume and data rate for both uplink and downlink.
Open-Source Enabled Beyond 5G Private Mobile Networks: From Concept to Prototype
2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC)
To enable emerging applications of fifth generation (5G) systems for vertical scenarios like mobi... more To enable emerging applications of fifth generation (5G) systems for vertical scenarios like mobile broadband and Internet of things (IoT) for vehicles and maritime sectors, a new trend towards softwarization and establishment of private networks is gaining momentum. Different from earlier generations of mobile systems where network functions are traditionally operated based on dedicated hardware, 5G infrastructures can be virtualized and services can be provided based on software, common hardware and servers. In this paper, we present an open-source based prototype private network which has been implemented in the framework of an EEA research project SOLID-B5G. The prototype network is a 5G non-standalone (NSA) network with a fourth generation (4G) evolved packet core (EPC) connecting both evolved NodeB (eNB) and next generation NodeB (gNB) to commercial off-the-shelf (COTS) user equipment (UE). The prototype is implemented based on two popular open-source software platforms, namely OAI RAN and srsRAN. Based on our implementations, extensive experiments have been conducted to assess the performance of the prototype network in terms of downlink/uplink throughput and latency between different parts of the system
Deploying Use Case Specific Network Slices Using An OSM Automation Platform
2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC)
Dynamic Offloading The SDN Control Plane In Large Area Networks By Condition-Aware Migration Of Forwarding Devices
2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC)
Distributed multi-controller deployments are explored in large SDN-controlled networks to achieve... more Distributed multi-controller deployments are explored in large SDN-controlled networks to achieve the control plane scalability and reliability. There are many types of research on the Controller Selection or Placement Problem (CSP/CPP). The majority of studies consider a static approach to optimize the controller placement and a static mapping of switches to controllers. However, in a dynamic topology (susceptible to controller overload, nodes, or link failure), the initial mapping of the forwarders-to-controllers and even the controller placement could become nonoptimal, w.r.t. QoS service requirements. The flow of data through multiple controllers could vary following an unequal distribution of the load between multiple controllers. To overcome the aforementioned challenge, a solution could be a limited dynamic switch migration if controller overload is detected. The contribution of this work-in-progress is a proposal of a powerful condition-aware mechanism for switch migration and its functionality validation. The system considered in this work include the Redis database, Ryu controller, Mininet, and Iperf and introduce the concept of Supervisor controller. Our prototype resolves multiple overloads, simultaneously fixing the load balancing problems within the SDN controller plane
2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC), 2022
Massive Internet of Things (mIoT) traffic in beyond fifth generation (B5G) networks is stimulatin... more Massive Internet of Things (mIoT) traffic in beyond fifth generation (B5G) networks is stimulating a surge of data volume and may cause potential performance degradation when multiple devices attempt to transmit at the same time. In this paper, we present three techniques that can mitigate the effects of concurrent transmissions of multiple devices and improve network performance. First, by considering cluster-formation through beamforming and device activity factor, the number of active devices in a cluster becomes much lower. Second, by enabling pseudo-orthogonal codes for initial access in addition to existing orthogonal codes, more devices are allowed to transmit simultaneously. Third, by deploying multiple-input multiple-output (MIMO) antennas at the base station, concurrent transmissions can be performed. Based on these techniques, network performance is assessed.
2022 International Conference on Software, Telecommunications and Computer Networks (SoftCOM)
The Software Defined Network (SDN) concept has changed the traditional network architecture to ac... more The Software Defined Network (SDN) concept has changed the traditional network architecture to achieve higher flexibility, the programmability of the network, the abstraction based on a logically centralized control plane, and the data (forwarding) plane decoupling. The SDN control centralization raises scalability issues. This work focuses on the scalability concerns of the Ryu SDN controller in various scenarios with distinct network topologies like tree and datacenter. Throughout this paper, we assess the consequences of distributing subscriber data requests over the controller performance into different network topologies also using a different number of subscribers. The performance of the Ryu is tested by observing the throughput of the controller. The assessment was accomplished by employing an Iperf traffic generator and Mininet. The paper studies two matters that impact the Ryu throughput: the exponential increase of the number of nodes in tree topology and the adoption of the same condition over a datacenter topology. The experiments involve the two types of topologies mentioned and a topologies size range from 200 to 2000 nodes and assessing the controller performance using the throughput as the performance criterion. The results of this research will be used in a research project oriented to 5G slicing, where the control plane is based on SDN.
Sensors, Aug 30, 2022
This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY
Sensors, 2022
IoT devices present an ever-growing domain with multiple applicability. This technology has favor... more IoT devices present an ever-growing domain with multiple applicability. This technology has favored and still favors many areas by creating critical infrastructures that are as profitable as possible. This paper presents a hierarchical architecture composed of different licensing entities that manage access to different resources within a network infrastructure. They are conducted on the basis of well-drawn policy rules. At the same time, the security side of these resources is also placed through a context awareness module. Together with this technology, IoT is used and Blockchain is enabled (for network consolidation, as well as the transparency with which to monitor the platform). The ultimate goal is to implement a secure and scalable security platform for the Smart Grid. The paper presents the work undertaken in the SealedGRID project and the steps taken for implementing security policies specifically tailored to the Smart Grid, based on advanced concepts such as Opinion Dynamics and Smart Grid-related Attribute-based Access Control.
Media content streaming and delivery is nowadays a high popularity service in Internet. Complex a... more Media content streaming and delivery is nowadays a high popularity service in Internet. Complex architectures like Content Delivery Networks, Content Aware/Oriented Networks have been proposed, where information/content objects are treated as first-class abstractions. As alternative, light architectures, are investigated and implemented, working on top of the current IP networking technologies. In both types of architectures, appropriate content server and path selection constitute the primary set of actions to be performed in the content delivery systems. Such a problem belongs to the general class of multicriteria optimization problem, having (in our case) as input, some information on servers, network and user context. This paper contains an extension of a preliminary work, focused on algorithms and policies for optimized paths and server selection. Simulation study results are presented here, to illustrate the better performance of multi-criteria optimization algorithms versus r...
This paper studies and evaluates the scalability properties of a resource management subsystem as... more This paper studies and evaluates the scalability properties of a resource management subsystem as a component of a networked media ecosystem. The overall system aims to offer multimedia delivery with configurable guarantees of Quality of Services, over multi-domain networks, to large communities of users. The transport infrastructure is based on creation of data plane logical slices named Virtual Content Aware Networks. These slices are realized under control of a management plane, comprising centralized per-network domain "controllers", cooperating to construct parallel data planes, which span multiple IP network domains independently managed. In this multi-domain context, the management system scalability properties are important, while the problem is similar to the multi-controller inter-communication in emerging multi-controller Software Defined Networking technologies. The management system architecture considered in this paper has been previously defined. This work a...
Uploads
Papers by Marius C Vochin
characteristics that cannot be captured by other types of neural networks. This is due to entries that are graphs that, by their nature, include, besides a certain topology (the spatial characteristic),
connections between nodes that model the costs (traffic load, speed, and road length) of the roads between nodes that can vary over time (the temporal characteristic). As a result, a prediction in a node
influences the prediction from adjacent nodes, and, globally, the prediction has more precision. On the other hand, an adequate neural network leads to a good prediction, but its complexity can be higher.
A recurrent neural network like LSTM is suitable for making predictions. A reduction in complexity can be achieved by choosing a relatively small number (usually determined by experiments) of hidden levels. The use of graphs as inputs to the neural network and the choice of a recurrent neural
network combined lead to good accuracy in traffic prediction with a low enough implementation effort that it can be accomplished on microcontrollers with relatively limited resources. The proposed
method minimizes the communication network (between vehicles and database servers) load and represents a reasonable trade-off between the communication network load and forecasting accuracy.
Traffic prediction leads to less-congested routes and, therefore, to a reduction in energy consumption.
The traffic is forecasted using an LSTM neural network with a regression layer. The inputs of the neural network are sequences—obtained from a graph that represents the road network—at specific moments in time that are read from traffic sensors or the outputs of the neural network
(forecasting sequences). The input sequences can be filtered to improve the forecasting accuracy. This general framework is based on the Contiki IoT operating system, which ensures support for wireless communication and the efficient implementation of processes in a resource-constrained system, and it is particularized to implement a graph neural network. Two cases are studied: one case in which the traffic sensors are periodically read and another case in which the traffic sensors are read when
their values’ changes are detected. A comparison between the cases is made, and the influence of filtering is evaluated. The obtained accuracy is very good and is very close to the accuracy obtained in an infinite precision simulation, the computation time is low enough, and the system can work in real time.
service providers must build and expand the use of edge computing. This can lead to
possible difficulties concerning how to place workloads of network services
automation and evaluate the costs in an optimal way. This paper’s intention is to
present a scenario with three different virtual network functions designed with
specific restrictions and different compute costs and the Control System Hub with
Open Source Mano solution to limit and face those restrictions. To achieve this, the
optimal placement algorithm uses the Control System Hub network analytic
platform along with virtual charmed factors for control and Quality-of-Service
measures.
communications (mMTC) devices covered by a fifth generation
(5G) cell could be huge, not all devices are always active. For
radio resource allocation, it is vital that a base station has the
knowledge on the number of active mMTC devices in each
frame. In this paper, we develop a mathematical framework to
perform pseudo-Bayesian estimation on the number of active
devices within a 5G new radio (NR) subframe. The estimation
is performed by the base station on a subframe-by-subframe
basis based upon which a transmission permission probability
is imposed on active devices for their transmissions in the next
subframe. The active devices include both backlogged devices
and new arrivals in the current subframe that will attempt for
medium access in the next subframe relying on a combination of
NR radio resources in both time and frequency domains. Such
an estimation constitutes an important component for the design
of grant-free access schemes for mMTC uplink traffic.