Use of Machine Learning for energy efficiency in present and future mobile networks

Sesto-Castilla, David; García-Villegas, Eduard; Lyberopoulos, George; Theodoropoulou, Eleni

doi:10.1109/wcnc.2019.8885478

Cited by 23 publications

(9 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on the results obtained using k -means clustering, we performed the data analysis. Compared with other clustering algorithms, k -means clustering is simple to implement and is suitable for low-dimensional data [ 36 , 37 ]. In contrast, the k -means clustering method requires prior knowledge about the optimal number of clusters [ 38 ], which is nearly impossible to achieve.…”

Section: Methodsmentioning

confidence: 99%

Cluster-Based Analysis of Infectious Disease Occurrences Using Tensor Decomposition: A Case Study of South Korea

Jung

Moon

Hwang

2020

IJERPH

View full text Add to dashboard Cite

For a long time, various epidemics, such as lower respiratory infections and diarrheal diseases, have caused serious social losses and costs. Various methods for analyzing infectious disease occurrences have been proposed for effective prevention and proactive response to reduce such losses and costs. However, the results of the occurrence analyses were limited because numerous factors affect the outbreak of infectious diseases and there are complex interactions between these factors. To alleviate this limitation, we propose a cluster-based analysis scheme of infectious disease occurrences that can discover commonalities or differences between clusters by grouping elements with similar occurrence patterns. To do this, we collect and preprocess infectious disease occurrence data according to time, region, and disease. Then, we construct a tensor for the data and apply Tucker decomposition to extract latent features in the dimensions of time, region, and disease. Based on these latent features, we conduct k-means clustering and analyze the results for each dimension. To demonstrate the effectiveness of this scheme, we conduct a case study on data from South Korea and report some of the results.

show abstract

Section: Methodsmentioning

confidence: 99%

Cluster-Based Analysis of Infectious Disease Occurrences Using Tensor Decomposition: A Case Study of South Korea

Jung

Moon

Hwang

2020

IJERPH

View full text Add to dashboard Cite

show abstract

“…On the other hand, the average computational time of the proposed algorithm is less than one minute (54.857 s). In modern HetNets, the traffic load of the network is monitored for longer time periods in the order of 5–15 min [ 6 , 64 , 65 , 66 ], while the on-off switching decisions of the SCs are usually taken every 15–60 min [ 64 , 66 ]. Thus, the proposed algorithm is suitable for practical implementation in real time.…”

Section: Proposed Algorithmsmentioning

confidence: 99%

Binary PSO with Classification Trees Algorithm for Enhancing Power Efficiency in 5G Networks

Osama

Ramly

Abdelhamid

2022

Sensors

View full text Add to dashboard Cite

The dense deployment of small cells (SCs) in the 5G heterogeneous networks (HetNets) fulfills the demand for vast connectivity and larger data rates. Unfortunately, the power efficiency (PE) of the network is reduced because of the elevated power consumption of the densely deployed SCs and the interference that arise between them. An approach to ameliorate the PE is proposed by switching off the redundant SCs using machine learning (ML) techniques while sustaining the quality of service (QoS) for each user. In this paper, a linearly increasing inertia weight–binary particle swarm optimization (IW-BPSO) algorithm for SC on/off switching is proposed to minimize the power consumption of the network. Moreover, a soft frequency reuse (SFR) algorithm is proposed using classification trees (CTs) to alleviate the interference and elevate the system throughput. The results show that the proposed algorithms outperform the other conventional algorithms, as they reduce the power consumption of the network and the interference among the SCs, ameliorating the total throughput and the PE of the system.

show abstract

“…are the key enablers of 5G to meet the promised level of QoS. Thus, implementing highly dense, complex and multi-layered 5G networks will require a higher degree of automation [81]. Although, according to [82], the existing 5G networks do not provide such level of flexibility or automation yet, Artificial Intelligence (AI) offer solutions to tackle the complexity of 5G and beyond [81] [82].…”

Section: Joint Spectrum Sharingmentioning

confidence: 99%

“…In wireless networks, by relating the system parameters to the desired objective [103], ML techniques can address the challenge of traditional optimization approaches, which tend to leave a large gap between the theoretical and real-time design of the network [104]. Therefore, for the development and automation of 5G networks, MLbased approaches are getting enormous attention and can be adopted in different aspects of cellular networks, e.g., interference management [104], beamforming [105][106][107], link quality estimation [108], 5G-based IoT [109], energy efficiency [81,110], resource management [111],…”

Section: Machine Learning In Wireless Networkmentioning

confidence: 99%

“…The central controller collects the predicted traffic requirements of the access network for each cell from ML agent and, after In [116], utilizing collected trace data, authors have developed a model to predict the ag-gregated traffic and, subsequently, reduce the monitoring effort. Classification and clustering of cells were used to improve the traffic predictions in [81,117]. However, these approaches require additional cell-level data.…”

Section: Machine Learning In Wireless Networkmentioning

confidence: 99%

See 1 more Smart Citation

Joint access-backhaul mechanisms in 5G cell-less architectures

Rony

View full text Add to dashboard Cite

Older generations of wireless networks, such as 1G and 2G were deployed using leased line, copper or fibre line as backhaul. Later, in 3G and 4G, microwave wireless links have also worked as backhaul links while the backbone of the network was still wireline-based. However, due to multiple different use cases and deployment scenarios of 5G, solo wireline based backhaul network is not a cost-efficient option for the operators anymore. For cost-efficient and fast deployment, wireless backhaul options are very attractive. As drawbacks, wireless backhaul links have capacity and distance limitations. To take the advantages of both the solutions, i.e., wired and wireless, 5G transport networks are anticipated to be a heterogeneous, complex, and with stringent performance requirements. To address the aforementioned challenges, wireless backhaul options are providing more attractive solutions, and hence, technologies using the same resources (e.g., frequency channels) may be used by both access and backhaul networks. In this scenario, blurring the separation line between access and backhaul networks allows resource sharing and cooperation between both the networks and minimizes the network deployment and maintenance cost significantly. Therefore, in 5G, the access and backhaul networks cannot be seen as separate entities; rather, we seek to integrate them together to ensure the best use of resources. In this thesis, firstly, we investigate the challenges and potential technologies of 5G transport network. Later, to address these challenges, we identify and present different approaches to perform joint access-backhaul mechanism. An initial performance evaluation of access-aware backhaul optimization is presented, where backhaul network is dynamically assigned with the required resources to serve the dynamic requirements of a 5G access network. The evaluation results and discussions manifest the resource efficiency of joint access-backhaul mechanisms. Functional splits in different layers of the access network comes as an intelligent solution to reduce the enormous capacity requirements of the transport network in a centralized radio access network approach, which tends to centralize almost all the functionalities into a central unit, leaving only radio frequency functions at the access points. From the joint access-backhaul mechanism perspective, we propose a novel technique, which takes the benefit of functional splits at physical layer, to design a heterogeneous transport network in an economical budget-limited and capacity-limited scenario. Till today, the limited capacity of the wireless backhaul links remains a challenge, and hence, frequency spectrum becomes scarce, and requires efficient utilization. To address this challenge, a joint spectrum sharing technique to implement joint accessbackhaul mechanism is presented. Evaluation results show that our proposed joint spectrum sharing technique, where spectrum allocation in the backhaul network follows the access network's traffic load, is fair and efficient in terms of spectrum utilization. We also propose a machine learning technique, which analyses data from a real network and estimates access network's traffic pattern, and subsequently, assigns bandwidth in the access network according to the traffic estimations. Presented evaluation results show that a well-trained machine learning model can be very efficient to obtain an efficient utilization of frequency spectrum. Las primeras generaciones de redes móviles, se implementaron utilizando líneas de cobre o fibra para la conexión entre la red de acceso y el núcleo de la red (conexión backhaul). Más tarde, los enlaces inalámbricos también han funcionado como backhaul mientras que la columna vertebral de la red seguía basada en cable. Sin embargo, debido a los múltiples escenarios de implementación de 5G, una red de backhaul basada solamente en cable ya no es una opción rentable para los operadores. Para una implementación rentable y rápida, las opciones de backhaul inalámbrico son muy atractivas. Como inconvenientes, los enlaces backhaul inalámbricos tienen limitaciones de capacidad y distancia. Para aprovechar las ventajas de ambas soluciones, es decir, cableadas e inalámbricas, se prevé que las redes de transporte 5G sean heterogéneas, complejas y con estrictos requisitos de rendimiento. Para abordar los desafíos antes mencionados, las opciones de backhaul inalámbrico brindan soluciones más atractivas y, por lo tanto, las tecnologías que usan los mismos recursos (por ejemplo, canales de frecuencia) pueden usarse tanto en las redes de acceso como en las de backhaul. En este escenario, desdibujar la línea de separación entre las redes de acceso y backhaul permite el intercambio de recursos y la cooperación entre ambas redes, y minimiza significativamente los costes de implementación y mantenimiento de la red. Por lo tanto, en 5G las redes de acceso y backhaul no pueden verse como entidades separadas; más bien consideraremos su integración para asegurar el mejor uso de los recursos. En esta tesis, en primer lugar, investigamos los desafíos y las tecnologías potenciales para la implementación de la red de backhaul 5G. Más tarde, para abordar dichos desafíos, identificamos diferentes enfoques para un mecanismo conjunto de gestión de la red de acceso y backhaul. Se presenta una evaluación de rendimiento inicial para la optimización de backhaul que tiene en cuenta el estado de la red de acceso, donde la red de backhaul se equipa dinámicamente con los recursos necesarios para cumplir con los requisitos de la red de acceso 5G. Los resultados de la evaluación manifiestan la mayor eficiencia de los mecanismos de gestión de recursos que consideran redes de acceso y backhaul conjuntamente. Las divisiones funcionales en diferentes capas de la red de acceso (functional splits) se presentan como una solución inteligente para reducir los enormes requisitos de capacidad de la red de transporte en un enfoque de red de acceso, que tiende a centralizar casi todas las funcionalidades en una unidad central, dejando solo las funciones más relacionadas con la transmisión/recepción de señales en los puntos de acceso. Desde la perspectiva del mecanismo conjunto de red de acceso y backhaul, proponemos una técnica novedosa, que aprovecha las divisiones funcionales en la capa física para diseñar una red de transporte heterogénea con un presupuesto económico y un escenario de capacidad limitada. Hasta el día de hoy, la capacidad limitada de los enlaces inalámbricos sigue siendo un desafío, dado que el espectro de frecuencias es escaso y requiere una utilización eficiente. Para hacer frente a este desafío, se presenta una técnica de gestión de recursos espectrales compartidos entre red de acceso y backhaul. Los resultados de la evaluación muestran que nuestra propuesta, donde la asignación de espectro en la red de backhaul se hace de acuerdo a la carga de tráfico de la red de acceso, es justa y eficiente. También proponemos una técnica de aprendizaje automático, que analiza datos de una red real y estima el patrón de tráfico de la red de acceso para, posteriormente, asignar ancho de banda en la red de acceso de acuerdo con dichas estimaciones. Los resultados de la evaluación presentados muestran que un modelo de aprendizaje automático bien entrenado puede ser una herramienta muy útil a la hora de obtener una utilización eficiente del espectro de frecuencias.

show abstract

Use of Machine Learning for energy efficiency in present and future mobile networks

Cited by 23 publications

References 7 publications

Cluster-Based Analysis of Infectious Disease Occurrences Using Tensor Decomposition: A Case Study of South Korea

Cluster-Based Analysis of Infectious Disease Occurrences Using Tensor Decomposition: A Case Study of South Korea

Binary PSO with Classification Trees Algorithm for Enhancing Power Efficiency in 5G Networks

Joint access-backhaul mechanisms in 5G cell-less architectures

Contact Info

Product

Resources

About