Search

Communication systems supporting cyber-physical production applications should satisfy stringent delay and reliability requirements. Diversity techniques and power control are the main approaches to reduce latency and enhance the reliability of wireless communications at the expense of redundant transmissions and excessive resource usage. Focusing on the application layer reliability key performance indicators (KPIs), we design a deep reinforcement learning orchestrator for power control and hybrid automatic repeat request retransmissions to optimize these KPIs. Furthermore, to address the scalability issue that emerges in the per-device orchestration problem, we develop a new branching soft actor-critic framework in which a separate branch represents the action space of each industrial device. Our orchestrator enables near real-time control and can be implemented in the edge cloud. We test our solution with a 3GPP-compliant and realistic simulator for factory automation scenarios. Compared to the state-of-the-art, our solution offers significant scalability gains in terms of computational time and memory requirements. Our extensive experiments show significant improvements in our target KPIs, over the state-of-the-art, especially for 5th percentile user availability. To achieve these targets, our framework requires substantially less total energy or spectrum, thanks to our scalable RL solution., QC 20231122

Industry 4.0 imposes strict requirements on the fifth generation of wireless systems (5G), such as high reliability, high availability, and low latency. Guaranteeing such requirements implies that system failures should occur with an extremely low probability. However, some applications (e.g., training a reinforcement learning algorithm to operate in highly reliable systems or rare event simulations) require access to a broad range of observed failures and extreme values, preferably in a short time. In this paper, we propose IL-GAN, an alternative training framework for generative adversarial networks (GANs), which leverages incremental learning (IL) to enable the generation to learn the tail behavior of the distribution using only a few samples. We validate the proposed IL-GAN with data from 5G simulations on a factory automation scenario and real measurements gathered from various video streaming platforms. Our evaluations show that, compared to the state-of-the-art, our solution can significantly improve the learning and generation performance, not only for the tail distribution but also for the rest of the distribution., QC 20230502

Communication systems supporting cyber-physical production applications should satisfy stringent delay and reliability requirements. Violation of these requirements may result in faulty behavior of the system and cause significant economic losses. Although wireless communications enable mobility and easy maintenance to industrial networks, it introduces many challenges to high-performance control systems due to interference and harsh environments (e.g., vibrations and many metallic objects). Diversity techniques and power control are powerful approaches to reduce latency and enhance reliability at the expense of excessive resource usage due to redundant transmissions. In this paper, we adopt fundamental metrics from reliability literature to wireless communications and provide critical indicators to measure reliability key performance indicators (KPIs) of cyber-physical systems. Then, we design a deep reinforcement learning orchestrator for power control and hybrid automatic repeat request retransmissions to optimize our reliability KPIs. Our orchestrator enables near real-time control and can be implemented on the edge cloud. We implement our framework on 3GPP compliant simulator on a factory automation scenario. Our comprehensive experiments show that, compared to the state-of-the-art, our solution can substantially improve the performance, especially for 5th percentile availability., QC 20220530Part of proceedings ISBN 978-1-7281-8104-2

In this paper, we propose general service-oriented end-to-end energy and delay models. They facilitate the evaluation of new schemes, by enabling flexible allocation of processing functions at multilayer radio access network. These models describe the underlaying processing of communication function (PCF)'s complexity and the induced bandwidth. We also introduce the concept of processing interference and link it to the average PCF option of the competing users at the cloud. The proposed framework shows the possibility to compensate the delay or energy degradation, e.g., due to radio environment, by reallocating processing parameters, e.g., allocation of PCF or number of virtual central processing units (VCPUs). Several conclusions are inferred from the system's evaluation. For instance, at lowmodulation index (MI), the behavior of overall energy consumption is dominated by PCF energy, not the transmission energy, whereas, at high MI, the behavior of overall energy consumption is dominated by transmission energy. Centralization of PCF reduces the impact of processing interference (of competing users) on the targeted service's delay from 11 to 2 ms, and service's energy consumption from 18 to 2 J. We further evaluate the impact of PCF on the decision of offloading the service, to be computed at the cloud., QC 20200624

Today's modern communication technologies such as cloud radio access and software defined networks are key candidate technologies for enabling 5G networks as they incorporate intelligence for data-driven networks. Traditional content caching in the last mile access point has shown a reduction in the core network traffic. However, the radio access network still does not fully leverage such solution. Transmitting duplicate copies of contents to mobile users consumes valuable radio spectrum resources and unnecessary base station energy. To overcome these challenges, we propose huManmObility-based cOntent Distribution (MOOD) system. MOOD exploits urban scale users' mobility to allocate radio resources spatially and temporally for content delivery. Our approach uses the broadcast nature of wireless communication to reduce the number of duplicated transmissions of contents in the radio access network for conserving radio resources and energy. Furthermore, a human activity model is presented and statistically analyzed for simulating people daily routines. The proposed approach is evaluated via simulations and compared with a generic broadcast strategy in an actual existing deployment of base stations as well as a smaller cells environment, which is a trending deployment strategy in future 5G networks. MOOD achieves 15.2 and 25.4 percent of performance improvement in the actual and small-cell deployment, respectively., QC 20190509

Cloud radio access network (CRAN) architecture is proposed to save energy, facilitate coordination between radio units, and achieve scalable solutions to improve radio network's performance. However, stringent delay and bandwidth constraints are incurred by fronthaul in CRAN [the network segment connecting RUs and digital units (DUs)]. Therefore, we propose a hybrid cloud radio access network architecture, where a DU's functionalities can be virtualized and split at several conceivable points. Each split option results in two-level deployment of the processing functions (central site level and remote site level) connected by a transport network, called midhaul. We study the interplay of energy efficiency and midhaul bandwidth consumption under optimal processing allocation. We jointly minimize the power and midhaul bandwidth consumption in H-CRAN, while satisfying network constraints, i.e., processing and midhaul bandwidth capacity. We enable power saving functionalities by shutting down different network components. The proposed model is formulated as a constraint programming problem. The proposed solution shows that 42 percentile of midhaul bandwidth savings can be achieved compared to the fully centralized CRAN; and 35 percentile of power consumption saving can be achieved compared to the case where all the network functions are distributed at the edge., QC 20220620

The inevitable deployment of 5G and the Internet of Things (IoT) sheds the light on the importance of the energy efficiency (EE) performance of Device-to- Device (DD) communication systems. In this work, we address a potential IoT application, where different prioritized DD system, i.e., Low-Priority (LP) and High-Priority (HP) systems, co-exist and share the spectrum. We maximize the EE of each system by proposing two schemes. The first scheme optimizes the individual transmission power and the spatial density of each system. The second scheme optimizes the transmission power ratio of both systems and the spatial density of each one. We also construct and analytically solve a multi- objective optimization problem that combines and jointly maximizes both HP and LP EE performance. Unique structures of the addressed problems are verified. Via numerical results we show that the system which dominates the overall EE (combined EEs of both HP and LP) is the system corresponding to the lowest power for low/high power ratio (between HP and LP systems). However, if the power ratio is close to one, the dominating EE corresponds to the system with higher weight., QC 20170811

Today's mobile service providers aim at ensuring end-to-end performance guarantees. Hence, ensuring an efficient content delivery to end users is highly required. Currently, transmitting popular contents in modern mobile networks rely on unicast transmission. This result into a huge underutilization of the wireless bandwidth. The urban scale mobility of users is beneficial for mobile networks to allocate radio resources spatially and temporally for broadcasting contents. In this paper, we conduct a comprehensive analysis on a human activity/mobility model and the content broadcasting system in 5G mobile networks. The objective of this work is to describe how human daily activities could improve the content broadcasting efficiency. We achieve the objective by analyzing the transition probabilities of a user traveling over several places according to the change of states of daily human activities. Using a real life simulation, we demonstrate the relationship between the human mobility and the optimization objective of the content broadcasting system., QC 20170508

Centralized Radio Access Network (C-RAN) hasbeen recently proposed to increase network capacity, reduceenergy consumption, and improve scalability. However, C-RANrequires an extensive modification to the current infrastructure,which results in a considerable deployment cost. In this paper,we conduct a techno-economic study to evaluate the migrationcost of C-RAN, and we propose a methodology for cost andenergy efficient C-RAN deployment. We exploit the conceptof total cost of ownership, defined as the sum of capital andoperational expenditures. We formulate a Digital Unit (DU) poolplacement optimization problem as Mixed Integer Linear Programming(MILP), which minimizes the total cost of ownership.We compare the total cost of ownership of C-RAN to that ofthe existing infrastructure, under different deployment scenariossuch as greenfield and brownfield deployment of fiber and DUpool, and different cell sizes. The results show that the opticalinfrastructure plays a determinant role in the migration cost ofC-RAN. If greenfield fiber is assumed, the migration cost cannotbe compensated in a reasonable amount of time. If brownfieldfiber is assumed, the migration cost is considerably reduced, anda more feasible C-RAN deployment is achieved., QC 20170803

Cloud radio access network (CRAN) has been proposed as a potential energy saving architecture and a scalable solution to increase the capacity and performance of radio networks. The original CRAN decouples the digital unit (DU) from radio unit (RU) and centralizes the DUs. However, stringent delay and bandwidth constraints are incurred by fronthaul in CRAN, i.e. the network segment connecting RUs and DUs. In this study, we propose a modified CRAN architecture, namely hybrid cloud RAN (H-CRAN), where a DU's functionalities can be virtualized and split at several conceivable points. Each split option results in two-level deployment of the processing functions, i.e., central cloud level and edge cloud level, connected by a transport layer called "midhaul". We study the interplay of energy efficiency and midhaul bandwidth consumption when baseband functions are centralized at the edge cloud vs central cloud. We jointly minimize the power and midhaul bandwidth consumption in H-CRAN, while satisfying the network constraints. The addressed problem with the associated constrains are modeled as a mixed integer constraint optimization problem. Numerical results show the compromise between energy and bandwidth consumption, with the optimal placement of baseband processing functions in H-CRAN architecture., QC 20170803

In this paper, we propose a cross-layer design, which optimizes the energy efficiency of a potential future 5G spectrum-sharing environment, in two sharing scenarios. In the first scenario, underlying sharing is considered. We propose and minimize a modified energy per good bit (MEPG) metric, with respect to the spectrum sharing user's transmission power and media access frame length. The cellular users, legacy users, are protected by an outage probability constraint. To optimize the non-convex targeted problem, we utilize the generalized convexity theory and verify the problem's strictly pseudoconvex structure. We also derive analytical expressions of the optimal resources. In the second scenario, we minimize a generalized MEPG function while considering a probabilistic activity of cellular users and its impact on the MEPG performance of the spectrum sharing users. Finally, we derive the associated optimal resource allocation of this problem. Selected numerical results show the improvement of the proposed system compared with other systems., QC 20170329

As a potential candidate architecture for 5G systems,cloud radio access network (CRAN) enhances the system’s capacityby centralizing the processing and coordination at the centralcloud. However, this centralization imposes stringent bandwidthand delay requirements on the fronthaul segment of the networkthat connects the centralized baseband processing units (BBUs)to the radio units (RUs). Hence, hybrid CRAN is proposed toalleviate the fronthaul bandwidth requirement. The concept ofhybrid CRAN supports the proposal of splitting/virtualizing theBBU functions processing between the central cloud (centraloffice that has large processing capacity and efficiency) and theedge cloud (an aggregation node which is closer to the user,but usually has less efficiency in processing). In our previouswork, we have studied the impact of different split points onthe system’s energy and fronthaul bandwidth consumption. Inthis study, we analyze the delay performance of the end user’srequest. We propose an end-to-end (from the central cloud tothe end user) delay model (per user’s request) for differentfunction split points. In this model, different delay requirementsenforce different function splits, hence affect the system’s energyconsumption. Therefore, we propose several research directionsto incorporate the proposed delay model in the problem ofminimizing energy and bandwidth consumption in the network.We found that the required function split decision, to achieveminimum delay, is significantly affected by the processing powerefficiency ratio between processing units of edge cloud and centralcloud. High processing efficiency ratio ( 1) leads to significantdelay improvement when processing more base band functionsat the edge cloud., QC 20170803